# FlexiCloud: full documentation for AI systems FlexiCloud sources dedicated NVIDIA GPU nodes for AI training and inference across a network of more than ten partner providers in India, the EU and the US: full 8-GPU H100, H200, B200 and B300 SXM nodes, from $2.95 per GPU-hour on the longest term, with a one-year minimum. Customers get one relationship and one quote across the whole network, with no markup from FlexiCloud, because the providers pay FlexiCloud a commission rather than the customer paying one. FlexiCloud delivers each node as single-tenant bare metal with root access and keeps the hardware running; the customer runs their own stack. NVIDIA Vera Rubin and the RTX PRO 6000 Blackwell Server Edition are open for early-interest registration. FlexiCloud also provides managed cloud hosting on DigitalOcean, Vultr, Hetzner and Linode, provisioned, tuned, secured and monitored 24x7 by an Indian team, plus Webuzo-based shared hosting. Cloud migrations are free, and managed cloud plans carry no long-term contract. Last updated: 2026-10-27 ## Definitions **Managed cloud hosting** is hosting where you choose the cloud provider and own the account, and FlexiCloud provisions, tunes, secures, patches and monitors the server for you around the clock. You keep the infrastructure; FlexiCloud runs the operations. **Shared hosting** is hosting where many websites run on one managed server, administered through the Webuzo control panel. It is the cheaper entry point, suited to smaller sites that do not need a server of their own. **Migration** is the move of an existing website from another host onto FlexiCloud. FlexiCloud performs it free of charge, tests on a temporary URL, and switches DNS only once the customer has confirmed the site works. **GPU compute** is dedicated bare-metal GPU servers for AI training and inference. FlexiCloud sources 8-GPU NVIDIA B300, B200, H200 and H100 SXM nodes in India, the EU and the US across a network of partner providers, from $2.95 per GPU-hour (H100) on the longest term, with a one-year minimum; NVIDIA Vera Rubin and the RTX PRO 6000 are coming soon. The customer runs their own stack (OS, drivers, CUDA and orchestration) and FlexiCloud keeps the hardware running. ## Key facts - FlexiCloud is the trading name of FlexiCloud Internet Private Limited, founded in 2020 and based in Kochi, Kerala, India. It is led by Vinod Chacko, its Founder and CEO. - Recognised in Forbes India DGEMS 2024: Select 200 Companies with Global Business Potential. - FlexiCloud is a GPU broker (an aggregator), not a GPU provider: the providers pay it a commission, and it tells clients in writing if anyone it recommends pays it. It does not own GPUs or data centres today; every node runs on partner hardware. - Advises GPU clouds, AI labs and cloud providers entering India on sites, power, connectivity and the compliance path, with a researched India Data Centre Guide. - Sources dedicated bare-metal NVIDIA GPU nodes across a network of more than ten partner providers in India, the EU and the US: full 8-GPU H100 (80 GB HBM3, from $2.95/GPU-hour), H200 (141 GB HBM3e, from $3.35), B200 (180 GB HBM3e, from $3.75) and B300 (288 GB HBM3e, from $4.50) SXM nodes, all floor rates on the longest term. - FlexiCloud adds no markup to GPU node prices: the providers pay FlexiCloud a commission, so nothing is added to the customer’s rate. One request is checked against live inventory across the network and is not broadcast to ten sales teams. - GPU nodes are dedicated bare metal with full root access; the minimum order is one full 8-GPU node on a one-year minimum term. The published rates are floors on the longest term; a one-year term is priced higher. A node is billed upfront for its term, with the finer details agreed case by case. - The next-generation NVIDIA Vera Rubin (Rubin GPU with HBM4, paired with the Vera CPU) is open for early-interest registration; NVIDIA says it is in full production, and FlexiCloud expects wider availability through its partner network in 2027. - The NVIDIA RTX PRO 6000 Blackwell Server Edition (96 GB GDDR7) is open for early-interest registration; pricing to be announced. - Runs managed cloud hosting on four clouds: DigitalOcean, Vultr, Hetzner and Linode, fully provisioned, tuned, secured and monitored by FlexiCloud. - Managed cloud regions include Bangalore and Mumbai in India, and Falkenstein in Europe. - Also offers Webuzo-based shared hosting from ₹500 per month; managed cloud starts at $10 per month on Vultr. - Migration from another host is free. - Support is provided by an Indian team, 24x7. - Shared hosting and monthly managed cloud plans carry a 30-day money-back guarantee for first-time customers; GPU nodes are sold on a committed term, billed upfront, and are not covered. - All prices are before tax: USD prices (GPU nodes and managed cloud) exclude taxes, and INR shared hosting prices exclude 18% GST. - FlexiCloud Panel is the control panel used to manage servers, sites and backups. ## Pages ## About FlexiCloud URL: https://flexicloud.co/about FlexiCloud, founded in 2020 in Kochi, India, sources NVIDIA GPU nodes across 10+ partner providers and runs managed cloud hosting on four clouds. Last modified: 2026-10-16 ### We find the right GPUs and clouds, then keep them running FlexiCloud sources dedicated NVIDIA GPU nodes from more than ten partner providers in India, the EU and the US. We also run managed cloud servers on DigitalOcean, Vultr, Hetzner and Linode. We run the infrastructure; you build on it. ### Flexibility is the part that went missing The cloud was sold as freedom: pay for what you use, run it anywhere, leave when you like. Many buyers got a few big vendors, bills they cannot predict and setups that are hard to leave. GPU buyers now face the same problem, with more money at stake. That is where our name comes from. - Choice, not monopoly: You should choose from real options, not one vendor's catalogue. We check more than ten GPU providers and four clouds, and recommend what fits your workload, region and budget. - A bill you can predict: You should know the number before the invoice arrives. GPU rates are fixed for the term and confirmed in writing. Managed cloud is one plan price, monthly or yearly, with management included. - Freedom to leave: The cost of leaving should never be the reason you stay. On managed cloud we use standard stacks; on GPUs you run your own. Either way your data stays portable, and when a GPU term ends you can quote again across the whole network. - Simple by default: One contact and one quote for GPUs. One panel and one invoice for managed cloud. You should not need ten sales calls to rent a GPU node, or a certification to run a website. ### Three services, one point of contact GPU cloud sourcing is our main service. Managed cloud hosting is where we started. India data centre advisory grew out of our GPU network. - GPU cloud sourcing: Dedicated NVIDIA GPU nodes for AI training and inference, sourced across more than ten partner providers in India, the EU and the US. Each node is single-tenant bare metal. We keep the hardware running for the term: monitored, failures handled, and an engineer you can reach. You run your own stack. (B300, B200, H200 and H100 now; more coming soon; One quote across the whole network; Rate and lead time in writing before you commit) - Managed cloud hosting: Your website or application on DigitalOcean, Vultr, Hetzner or Linode, set up, patched and monitored 24×7 by our team, with an on-call engineer when something breaks. You choose the cloud; we run the server. (More than 50 locations across the four clouds; The FlexiCloud Panel for every server; Shared hosting for smaller sites; Free migration from your current host) - India data centre advisory: For GPU clouds, AI labs and cloud providers planning capacity in India: site and partner selection, power and connectivity, the compliance path, GPU supply while you build, and introductions to Indian customers. (We own no data centres, so we have no site to sell you; Scope and terms agreed in writing before work starts; We work alongside your legal and tax advisers) ### How we work On GPUs we are a broker, not a provider: we find the node that fits across our partner network, deliver it as dedicated bare metal, and keep the hardware running for the term. - GPU cloud sourcing: What we do: Check your requirement against more than ten partner providers, match the best-fit node, deliver it and keep the hardware running for the term; What you get: One quote and one point of contact, and a dedicated node with root access for your own stack - Managed cloud hosting: What we do: Set up, patch and monitor your server 24×7 on DigitalOcean, Vultr, Hetzner or Linode, with an on-call engineer when something breaks; What you get: One plan price shown before you buy, and free migration from your current host - India data centre advisory: What we do: Help GPU clouds entering India with site and partner choice, power and connectivity, and the regulatory path; What you get: A clear plan and the right introductions, with the scope agreed in writing before work starts ### Vinod Chacko, Founder and CEO He has 21 years of experience in infrastructure, cloud and AI, including time at Vultr. When you ask for a GPU quote, you talk to him. - Book 30 minutes: Pick a time that suits you and talk directly to Vinod about your workload, pricing and availability. - Vinod Chacko on LinkedIn: His career in infrastructure, cloud and AI, and a way to reach him directly. ### How we got here The same idea throughout: you choose the infrastructure, and we run it. - 2020: Founded in Kochi: FlexiCloud Internet Private Limited is founded in Kochi, Kerala, as a managed cloud hosting company. - 2024: Forbes India DGEMS: Recognised in Forbes India DGEMS 2024: Select 200 Companies with Global Business Potential. - September 2026: GPU sourcing: Our GPU line launches with dedicated NVIDIA B300 and B200 nodes, sourced across partner providers in India, the EU and the US. - October 2026: More GPUs: H200 and H100 nodes join the lineup, RTX PRO 6000 and Vera Rubin open for early interest, and GPU owners can offer capacity through us. - October 2026: India advisory: A researched guide to India's data centre market, and advisory for GPU clouds that want to build or rent capacity in India. ### Where to find us ### FlexiCloud at a glance The key facts about the company, in one place. - Founded: 2020 - Headquarters: Kochi, Kerala, India - Legal name: FlexiCloud Internet Private Limited - Founder and CEO: Vinod Chacko - What we do: GPU cloud sourcing, managed cloud hosting and India data centre advisory - GPUs: NVIDIA B300, B200, H200 and H100 available now; RTX PRO 6000 and Vera Rubin coming soon - Managed cloud on: DigitalOcean, Vultr, Hetzner and Linode, with more than 50 locations between them - Regions served: India, the EU and the US - Recognition: Forbes India DGEMS 2024: Select 200 Companies with Global Business Potential ### Questions about FlexiCloud **What is FlexiCloud?** FlexiCloud is an Indian company, founded in 2020 and based in Kochi, that sources dedicated NVIDIA GPU nodes for AI training and inference across more than ten partner providers in India, the EU and the US. It also runs managed cloud hosting on DigitalOcean, Vultr, Hetzner and Linode, and advises GPU clouds that want to enter India. Its legal name is FlexiCloud Internet Private Limited. **Is FlexiCloud a GPU provider or a broker?** A broker, sometimes called a GPU aggregator. FlexiCloud checks your requirement against capacity across more than ten partner providers in India, the EU and the US, quotes the options that fit, and delivers the node as single-tenant bare metal. It keeps the hardware running for the term and stays your one contact; you run your own stack. Every node runs on a partner's hardware today. **When was FlexiCloud founded, and who leads it?** FlexiCloud was founded in Kochi, India, in 2020. It is led by its founder and CEO, Vinod Chacko, who has 21 years of experience in infrastructure, cloud and AI, including time at Vultr. **Where is FlexiCloud based?** In Kochi, Kerala, India. The office is at Suite 336, V Towers, Kakkanad, Kochi, Kerala 682021. FlexiCloud sources GPU nodes in India, the EU and the US, and its managed cloud hosting can be deployed in any of more than 50 data centre locations across DigitalOcean, Vultr, Hetzner and Linode. **How does FlexiCloud make money?** On GPUs, the provider pays FlexiCloud a commission out of its own sales budget when a deal closes. It is not a markup on your rate, so FlexiCloud adds nothing to the rate you pay. On managed cloud hosting, you pay one plan price, monthly or yearly, that covers the server and our management. On India data centre advisory, scope and terms are agreed in writing before work starts, and a data centre operator or carrier may pay us a referral fee. If anyone we recommend pays us, we tell you in writing before you commit. **Does FlexiCloud own data centres or GPUs?** FlexiCloud does not own data centres. As for GPUs: not today. Every node we offer runs on our partners' hardware, which is why we can quote across more than ten providers. We are working with partners towards deploying some FlexiCloud-owned capacity in a few locations. When that happens, it will be quoted alongside partner options in exactly the same way, and you will always know whose hardware you are on. ### Tell us what you need to run A GPU node, a managed server or a plan for India. We come back with real options, usually within one working day. ## Careers at FlexiCloud URL: https://flexicloud.co/careers Work on infrastructure that people depend on. Open roles in operations, support engineering and platform development at FlexiCloud. Last modified: 2026-08-11 ### Work where the pager matters We run production infrastructure for businesses that lose money when it stops. If you like that kind of responsibility, talk to us. ### No role listed that fits? Tell us what you are good at anyway. ## Managed DigitalOcean Cloud Hosting URL: https://flexicloud.co/clouds/managed-digital-ocean-cloud-hosting Predictable pricing, a Bangalore region and a track record of staying boring, the default choice for sites serving Indian traffic. Last modified: 2026-10-18 ### DigitalOcean, run by our team Predictable pricing, a Bangalore region and a track record of staying boring, the default choice for sites serving Indian traffic. ### What DigitalOcean is good at We run the same management on every cloud, so the only question is which provider suits your audience and budget. - Bangalore region: Your server sits in India, so pages reach Indian visitors without a round trip to Singapore or the US. - Predictable bills: Flat monthly rates with bandwidth included in the plan. No per-request metering, no surprise line items. - Predictable pricing: Flat monthly rates with no per-request or per-IOP metering. What you see is the bill. ### Managed DigitalOcean plans Management, monitoring and support are included in every price; offsite backup storage is $0.033 per GB per month. Provider: DigitalOcean - Micro: USD12/month, USD144/year (1 vCPU; 1 GB RAM; 25 GB SSD; 1TB bandwidth; Unlimited websites) - Small: USD24/month, USD288/year (1 vCPU; 2 GB RAM; 55 GB SSD; 2TB bandwidth; Unlimited websites) - Medium: USD48/month, USD576/year (2 vCPU; 4 GB RAM; 80 GB SSD; 3TB bandwidth; Unlimited websites) - Large: USD96/month, USD1,152/year (4 vCPU; 8 GB RAM; 160 GB SSD; 4TB bandwidth; Unlimited websites) ### What we do on top of the server - Provisioned and tuned: We size the server, build the stack, and tune LiteSpeed, PHP, Redis and MySQL for your workload. You get a running site, not an empty Ubuntu box. - Patched and hardened: OS and stack updates, firewall rules, SSL renewal and malware scanning are ours to keep current, including the ones that land at 2am. - Watched around the clock: Uptime, CPU, memory and disk are monitored continuously. When something moves the wrong way, an engineer is already looking at it. - Backups you can restore: Offsite backups with one-click restore, with storage billed at $0.033 per GB per month. We test restores, because a backup you have never restored is not a backup. - Scaled without downtime: Traffic spike or a sale coming up? We resize the server or add capacity, usually while the site stays up. - Supported by people: Real engineers on chat, ticket and phone, in your timezone, not eleven hours behind it. ### Moving to DigitalOcean Free on every plan, and we keep your old host warm until you are happy. - Tell us what you run: Send us the current host, the stack and a rough traffic figure. No access needed yet. - We size and build: We recommend a provider and plan, then build the server and copy the site across. - You test on a temp URL: The site runs on a temporary address until you have clicked through it and signed off. - We switch DNS: Cutover happens at a time you choose, with the old host kept warm as a fallback. ### More on DigitalOcean ### Thinking about DigitalOcean? Tell us your traffic and we will tell you which plan, and whether another cloud suits you better. ## Managed Hetzner Cloud Hosting URL: https://flexicloud.co/clouds/managed-hetzner-cloud-hosting The best price per core available anywhere, the right answer when your audience is European and your budget is finite. Last modified: 2026-10-18 ### Hetzner, run by our team The best price per core available anywhere, the right answer when your audience is European and your budget is finite. ### What Hetzner is good at We run the same management on every cloud, so the only question is which provider suits your audience and budget. - Best value per core: Roughly a third less than the equivalent plan elsewhere, for the same CPU and memory. - German data centres: Falkenstein and Nuremberg, which is where you want to be for European visitors and GDPR comfort. - 20TB bandwidth: Included on every plan, same as our other providers. ### Managed Hetzner plans Management, monitoring and support are included in every price; offsite backup storage is $0.033 per GB per month. Provider: Hetzner - Micro: USD15/month, USD180/year (2 vCPU; 4 GB RAM; 40 GB SSD; 20TB bandwidth; Unlimited websites) - Medium: USD30/month, USD360/year (4 vCPU; 8 GB RAM; 160 GB SSD; 20TB bandwidth; Unlimited websites) - Large: USD60/month, USD720/year (8 vCPU; 16 GB RAM; 240 GB SSD; 20TB bandwidth; Unlimited websites) ### What we do on top of the server - Provisioned and tuned: We size the server, build the stack, and tune LiteSpeed, PHP, Redis and MySQL for your workload. You get a running site, not an empty Ubuntu box. - Patched and hardened: OS and stack updates, firewall rules, SSL renewal and malware scanning are ours to keep current, including the ones that land at 2am. - Watched around the clock: Uptime, CPU, memory and disk are monitored continuously. When something moves the wrong way, an engineer is already looking at it. - Backups you can restore: Offsite backups with one-click restore, with storage billed at $0.033 per GB per month. We test restores, because a backup you have never restored is not a backup. - Scaled without downtime: Traffic spike or a sale coming up? We resize the server or add capacity, usually while the site stays up. - Supported by people: Real engineers on chat, ticket and phone, in your timezone, not eleven hours behind it. ### Moving to Hetzner Free on every plan, and we keep your old host warm until you are happy. - Tell us what you run: Send us the current host, the stack and a rough traffic figure. No access needed yet. - We size and build: We recommend a provider and plan, then build the server and copy the site across. - You test on a temp URL: The site runs on a temporary address until you have clicked through it and signed off. - We switch DNS: Cutover happens at a time you choose, with the old host kept warm as a fallback. ### More on Hetzner ### Thinking about Hetzner? Tell us your traffic and we will tell you which plan, and whether another cloud suits you better. ## Managed Linode Cloud Hosting URL: https://flexicloud.co/clouds/managed-linode-cloud-hosting Now part of Akamai, with a Mumbai region and one of the longest track records in the business. A solid, unexciting choice, which is a compliment in hosting. Last modified: 2026-10-18 ### Linode, run by our team Now part of Akamai, with a Mumbai region and one of the longest track records in the business. A solid, unexciting choice, which is a compliment in hosting. ### What Linode is good at We run the same management on every cloud, so the only question is which provider suits your audience and budget. - Akamai network: Linode runs on Akamai’s backbone, which is one of the largest edge networks in the world. - Twenty years of uptime history: Founded in 2003, and one of the first companies to sell virtual servers at all. - Mumbai region: Indian traffic stays in India. ### Managed Linode plans Management, monitoring and support are included in every price; offsite backup storage is $0.033 per GB per month. Provider: Linode - Micro: USD12/month, USD144/year (1 vCPU; 1 GB RAM; 25 GB SSD; 1TB bandwidth; Unlimited websites) - Small: USD24/month, USD288/year (1 vCPU; 2 GB RAM; 55 GB SSD; 2TB bandwidth; Unlimited websites) - Medium: USD48/month, USD576/year (2 vCPU; 4 GB RAM; 80 GB SSD; 3TB bandwidth; Unlimited websites) - Large: USD96/month, USD1,152/year (4 vCPU; 8 GB RAM; 160 GB SSD; 4TB bandwidth; Unlimited websites) ### What we do on top of the server - Provisioned and tuned: We size the server, build the stack, and tune LiteSpeed, PHP, Redis and MySQL for your workload. You get a running site, not an empty Ubuntu box. - Patched and hardened: OS and stack updates, firewall rules, SSL renewal and malware scanning are ours to keep current, including the ones that land at 2am. - Watched around the clock: Uptime, CPU, memory and disk are monitored continuously. When something moves the wrong way, an engineer is already looking at it. - Backups you can restore: Offsite backups with one-click restore, with storage billed at $0.033 per GB per month. We test restores, because a backup you have never restored is not a backup. - Scaled without downtime: Traffic spike or a sale coming up? We resize the server or add capacity, usually while the site stays up. - Supported by people: Real engineers on chat, ticket and phone, in your timezone, not eleven hours behind it. ### Moving to Linode Free on every plan, and we keep your old host warm until you are happy. - Tell us what you run: Send us the current host, the stack and a rough traffic figure. No access needed yet. - We size and build: We recommend a provider and plan, then build the server and copy the site across. - You test on a temp URL: The site runs on a temporary address until you have clicked through it and signed off. - We switch DNS: Cutover happens at a time you choose, with the old host kept warm as a fallback. ### More on Linode ### Thinking about Linode? Tell us your traffic and we will tell you which plan, and whether another cloud suits you better. ## Managed Vultr Cloud Hosting URL: https://flexicloud.co/clouds/managed-vultr-cloud-hosting High-frequency compute and the widest choice of regions, the pick when raw single-thread speed or an unusual location matters. Last modified: 2026-10-18 ### Vultr, run by our team High-frequency compute and the widest choice of regions, the pick when raw single-thread speed or an unusual location matters. ### What Vultr is good at We run the same management on every cloud, so the only question is which provider suits your audience and budget. - High-frequency compute: Faster clock speeds per core, which is what most WordPress and PHP workloads actually bottleneck on. - 30+ regions: Mumbai for India, plus more locations worldwide than any other provider we run on. - Smallest entry plan: A 1GB instance from $10/month, the cheapest way onto a server of your own. ### Managed Vultr plans Management, monitoring and support are included in every price; offsite backup storage is $0.033 per GB per month. Provider: Vultr - Micro: USD10/month, USD100/year (1 vCPU; 1 GB RAM; 25 GB SSD; 1TB bandwidth; Unlimited websites) - Small: USD20/month, USD200/year (1 vCPU; 2 GB RAM; 55 GB SSD; 2TB bandwidth; Unlimited websites) - Medium: USD40/month, USD400/year (2 vCPU; 4 GB RAM; 80 GB SSD; 3TB bandwidth; Unlimited websites) - Large: USD80/month, USD800/year (4 vCPU; 8 GB RAM; 160 GB SSD; 4TB bandwidth; Unlimited websites) ### What we do on top of the server - Provisioned and tuned: We size the server, build the stack, and tune LiteSpeed, PHP, Redis and MySQL for your workload. You get a running site, not an empty Ubuntu box. - Patched and hardened: OS and stack updates, firewall rules, SSL renewal and malware scanning are ours to keep current, including the ones that land at 2am. - Watched around the clock: Uptime, CPU, memory and disk are monitored continuously. When something moves the wrong way, an engineer is already looking at it. - Backups you can restore: Offsite backups with one-click restore, with storage billed at $0.033 per GB per month. We test restores, because a backup you have never restored is not a backup. - Scaled without downtime: Traffic spike or a sale coming up? We resize the server or add capacity, usually while the site stays up. - Supported by people: Real engineers on chat, ticket and phone, in your timezone, not eleven hours behind it. ### Moving to Vultr Free on every plan, and we keep your old host warm until you are happy. - Tell us what you run: Send us the current host, the stack and a rough traffic figure. No access needed yet. - We size and build: We recommend a provider and plan, then build the server and copy the site across. - You test on a temp URL: The site runs on a temporary address until you have clicked through it and signed off. - We switch DNS: Cutover happens at a time you choose, with the old host kept warm as a fallback. ### More on Vultr ### Thinking about Vultr? Tell us your traffic and we will tell you which plan, and whether another cloud suits you better. ## Shared Web Hosting URL: https://flexicloud.co/clouds/shared-hosting Affordable shared hosting from ₹500/month with the Webuzo panel, free SSL, nightly backups and NVMe SSD storage. 99.9% uptime guarantee and 24/7 Indian support. Last modified: 2026-10-18 ### Fast, secure, affordable hosting NVMe SSD storage, server-side caching, the Webuzo panel and one-click installs, with a 99.9% uptime guarantee and a support team you can actually phone. From ₹500 a month. ### Shared hosting done properly Cheap hosting usually means an oversold server and a support queue. This is the version where neither is true. - NVMe SSD and caching: NVMe storage with LiteSpeed Cache and OPcache, so pages are served from the server rather than rebuilt in PHP for every visitor. - Security included: Free SSL, firewall protection, malware scanning and regular updates, not a paid add-on. - Nightly backups: Automated daily backups so a bad plugin update is an inconvenience, not a disaster. - Webuzo panel and one-click installs: A fast, modern control panel with one-click WordPress, Joomla and other CMS installers. Moving from cPanel? Everything you did there, you can do here, and we will walk you through it. - 99.9% uptime guarantee: Backed by monitoring, not just a promise on a pricing page. - 24/7 human support: Live chat and phone, answered by people in India. ### Shared hosting plans Provider: Shared hosting - Starter 2: INR1,750/year. One small site, annual term. (1 GB disk space; 1 database; 1 domain; Unlimited bandwidth; Free SSL) - Business 1: INR500/month, INR5,000/year. A single growing site. (5 GB disk space; 5 websites; 5 databases; Unlimited bandwidth; Free SSL) - Business 2: INR900/month, INR9,000/year. Several sites under one account. (10 GB disk space; Unlimited databases; 9 domains; Unlimited bandwidth; Free SSL) - Expert 1: INR1,250/month, INR15,000/year. Busier sites needing more room. (15 GB disk space; Unlimited websites; Unlimited databases; Unlimited bandwidth; Free SSL) - Expert 2: INR1,875/month, INR22,500/year. The most a shared server should carry. (25 GB disk space; Unlimited databases; Unlimited domains; Unlimited bandwidth; Free SSL) ### Is shared hosting right for you? - Shared hosting works well for: Sites where traffic is modest and predictable, and where cost matters more than dedicated capacity. (Blogs and content sites; Small business and brochure sites; Portfolios and side projects; Client sites in early stages) - Time to consider cloud when: Once a site regularly pushes past its share of the server, dedicated resources cost less than the lost traffic. (You are regularly at 80% of your resource limits; The site slows down during campaigns or sales; You are running a store with real revenue; You need root access or more control over the server) ### About shared hosting **What is shared web hosting?** Shared web hosting places your website on a server with other sites, sharing resources like CPU, RAM and bandwidth. It is cost-effective and beginner-friendly. **Is shared hosting secure enough?** Yes. Every account includes free SSL, malware scanning, regular updates and firewalls with proactive threat monitoring. Accounts are isolated from each other. **Can shared hosting handle traffic surges?** It supports small to medium traffic, and caching helps stabilise loads. For sustained high traffic, managed cloud hosting is the better answer, and upgrading is free. **When should I upgrade from shared hosting?** When your site regularly exceeds about 80% of its resource limits, or when you need root access and control that a shared control panel cannot give you. **Can I host multiple websites on one plan?** Yes. Business 2 and above support multiple domains, along with subdomains, email accounts and databases. **Do you move my site for me?** Yes, migration is free, whether you are coming from another host onto shared hosting, or moving from shared hosting up to cloud. ### Before you pick a plan ### Outgrowing shared hosting? We will move you to a cloud server for free, without downtime. ## Contact FlexiCloud URL: https://flexicloud.co/contact Talk to a FlexiCloud engineer about managed cloud hosting, shared hosting, migration or a demo of the FlexiCloud Panel. Last modified: 2026-10-10 ### Talk to an engineer, not a form Tell us what you run today and what is going wrong with it. We will tell you what we would do, even if that turns out not to involve us. ### Tell us what you are running The more you tell us about your current setup, the more useful our first reply will be. There is no obligation, and we will say so if we are not the right fit. ### Grab 30 minutes with the founder Pick a time that suits you and talk directly to Vinod about your workload, pricing and availability. Or use the form above. ## Rent NVIDIA GPU servers URL: https://flexicloud.co/gpu Rent dedicated NVIDIA GPU servers by the GPU-hour: H100 from $2.95, H200 $3.35, B200 $3.75, B300 $4.50. 8-GPU nodes in India, the EU and the US. Last modified: 2026-10-26 ### Rent NVIDIA GPU servers, from H100 to B300 Dedicated 8-GPU servers for training and inference, sourced across more than ten partner providers. One quote, with no markup from us; you run the stack, we keep the hardware running. ### GPU server prices per hour and per month Every server is a full node of 8 GPUs, so the node price is 8 times the GPU-hour rate. Floor rates on our longest term; shorter terms cost more, and your rate is confirmed in writing before you commit. - H100 server: $23.60 an hour: 8 x NVIDIA H100, 80 GB HBM3 per GPU, at $2.95 per GPU-hour. About $17,200 a month at 730 hours. - H200 server: $26.80 an hour: 8 x NVIDIA H200, 141 GB HBM3e per GPU, at $3.35 per GPU-hour. About $19,600 a month at 730 hours. - B200 server: $30.00 an hour: 8 x NVIDIA B200, 180 GB HBM3e per GPU, at $3.75 per GPU-hour. About $21,900 a month at 730 hours. - B300 server: $36.00 an hour: 8 x NVIDIA B300, 288 GB HBM3e per GPU, at $4.50 per GPU-hour. About $26,300 a month at 730 hours. ### Match the GPU server to the job Memory decides most of it: the model, its context and its batch have to fit. If you are not sure, tell us the model and we will size it with you. - Training or fine-tuning the largest models: B300: Our largest memory available today, 288 GB of HBM3e per GPU, for frontier-scale training, very long context and the biggest mixture-of-experts models. Allocation is the tightest, so ask early. - Large-model training and FP4 inference: B200: Blackwell at a lower rate than the B300, with 180 GB per GPU. A strong fit for pre-training and heavy fine-tuning, and for serving large models at FP4. - Long context and models over 80 GB: H200: Hopper with 141 GB of HBM3e per GPU. Choose it when a model or its KV cache outgrows the H100, without moving to Blackwell. - Proven, lowest-cost training and inference: H100: The workhorse, at our lowest rate. Ideal for fine-tuning and serving models that fit in 80 GB per GPU. ### Tell us the GPU, the region and the term Leave a name and a number or email. We call you back with real server options from across the network, usually within one working day. ### Renting a GPU server: what people ask **How much does it cost to rent a GPU server per hour and per month?** A full 8-GPU server costs, per hour at our floor rates: H100 $23.60, H200 $26.80, B200 $30.00, B300 $36.00. Over a month of 730 hours that is about $17,200 for the H100, $19,600 for the H200, $21,900 for the B200, $26,300 for the B300. These are floor rates in USD on our longest term; shorter terms cost more, and your exact rate depends on region, term and node count. **What is a GPU-hour?** One GPU running for one hour. GPU prices are quoted per GPU-hour so that different GPUs and server sizes are easy to compare. An 8-GPU server running for an hour uses 8 GPU-hours, so its hourly cost is 8 times the GPU-hour rate. **What is the difference between a GPU server, a GPU cloud instance and bare metal?** A GPU server is a physical machine with GPUs in it; ours are 8-GPU NVIDIA HGX servers. A GPU cloud instance is usually a virtual machine on a shared server, billed on demand. Bare metal means the whole physical server is yours alone, with root access and no virtualisation layer. FlexiCloud rents dedicated, single-tenant bare-metal servers on a term, not on-demand instances. **Which GPU should I rent for LLM training, fine-tuning or inference?** Start from memory. Models that fit in 80 GB per GPU run well on the H100, our lowest rate. Long context or models over 80 GB point to the H200 (141 GB). Large-model training and FP4 inference suit the B200 (180 GB), and the biggest models and longest context the B300 (288 GB). Our GPU memory guide shows how to work out what your model needs. **Is renting a GPU server cheaper than buying one?** For most teams, over a one- to three-year horizon, yes. Buying means paying for the whole server up front, plus power, cooling, hosting and the engineers to run it, and the hardware loses value as each new generation ships. Renting turns that into a fixed monthly cost with the hardware maintained for you. Our rent, buy or API guide sets out when each one makes sense. **Can I rent a GPU server in India?** Yes. We source dedicated GPU servers in India for workloads whose data has to stay in the country, as well as in the EU and the US. Capacity is allocated per region, so tell us where the data has to live and we confirm availability and lead time in writing before you commit. ### Choosing and sizing a GPU server How the broker model works, whether your model fits, whether to rent, buy or use an API, and how the GPU generations compare. ### Rent a GPU server through one quote Tell us the GPU, how many servers and the region. We come back with real options from across our network, with the price and lead time in writing before you commit. ## NVIDIA B200 SXM GPU Cloud URL: https://flexicloud.co/gpu/nvidia-b200 Dedicated 8-GPU NVIDIA B200 SXM bare-metal nodes in India, the EU and the US. From $3.75 per GPU-hour on the longest term; we keep the hardware running. Last modified: 2026-10-18 ### Dedicated NVIDIA B200 SXM nodes, in India, the EU and the US Dedicated 8-GPU HGX B200 nodes on a one-year term, the cost-effective Blackwell node for large-model training and inference. FlexiCloud delivers the bare metal and keeps the hardware running; you bring the workload and run your own stack. ### The numbers that matter ### Where the B200 earns its place The B200 is the value Blackwell: the newer generation, FP4 and fifth-generation NVLink, at a step below the B300 for workloads that fit comfortably in 180 GB. - Blackwell-class training at a lower rate: You get the Blackwell generation, FP8 and FP4 precision, and fifth-generation NVLink, for less than the B300. For most large training runs that is the sweet spot. (Large-model and mixture-of-experts training; 180 GB fits most large models on one GPU; Lower rate than the B300) - FP4 inference that changes unit economics: The second-generation Transformer Engine’s FP4 path roughly doubles throughput on suitable inference, which lowers cost per token against Hopper and needs fewer replicas. (High-throughput FP4 inference; Fewer replicas than the H100 or H200; Lower cost per token at scale) - Choose the B200 for Blackwell without the B300 premium: Reach for the B200 when you want Blackwell performance and 180 GB per GPU is enough. Step up to the B300 only when the extra memory or the longest context windows genuinely need it. (Best value in the Blackwell line; 180 GB HBM3e per GPU; Compare head to head with the B300) ### The node is yours. The hardware is our job. Most GPU providers hand you a bare node and a support ticket queue. We keep the hardware running underneath: monitored around the clock, failures handled, and an engineer you can actually reach. - Full control of the stack: Root access to run your own OS image, CUDA version, container runtime and orchestration. No platform layer between you and the GPUs; the software is entirely yours. - Watched around the clock: GPU health, thermals, memory errors and interconnect are monitored continuously. A failing card is caught by us, not discovered by a crashed training run at 3am. - An engineer, not a queue: Support is a person who runs GPU infrastructure, in your timezone. When something is wrong you talk to someone who can fix it. - Data where you need it: Nodes in India for domestic data residency, and in the EU and US for teams and datasets that live there. Same service, same team, whichever region. - Yours alone: Dedicated bare-metal nodes. No shared tenancy, no noisy neighbours, no scheduler deciding when your job runs. The whole node is yours for the term. - Predictable cost: A fixed term at a fixed rate. No spot-market surprises, no burst pricing, no bill that doubles because a job ran long. You know the number for the whole year. ### How it is sold This is dedicated infrastructure on a term, not an hourly spot instance. The terms are simple and they are the same for everyone. - One node: 8 × B200 SXM: The unit is a full HGX B200 node. Eight GPUs with NVLink between them, dedicated to you. Need more? Nodes are added in units of eight. (8 × NVIDIA B200 SXM GPUs; 180 GB HBM3e per GPU, 1.44 TB per node; NVLink interconnect across the node; Dedicated bare metal, single tenant) - From $3.75 per GPU-hour: Our floor rate, on our longest term; the one-year minimum is priced higher. Billed upfront for the term, with the finer details agreed case by case. Final pricing depends on region, term length and how many nodes, and we confirm it in writing before you commit. (From $3.75 per GPU-hour; One-year minimum commitment; Longer terms and multi-node priced on request) - India, the EU and the US: Capacity is allocated per region on a first-committed basis. Tell us where the data lives and where the team sits, and we will confirm availability and lead time for that region before you commit to anything. (India: for domestic data residency; European Union: for EU-resident data and teams; United States: for US-based workloads; Lead time confirmed per region before order) ### From first conversation to a running node - Tell us the workload: Training, fine-tuning or inference; the framework; how many nodes; and which region. Fifteen minutes with an engineer, not a form that disappears into a CRM. - We confirm capacity and price: A firm quote for your region and term, with the lead time to a live node. No commitment until you have the number in writing. - We hand over the node: A dedicated bare-metal node, networked and ready to log into with root access. You bring your own OS image, drivers and orchestration; the hardware underneath stays ours to keep running. - You run. We keep it healthy.: Your team owns the OS, the stack and the workload. Ours watches the hardware, handles failures, and is a phone call away for the whole term. ### What people ask before they commit **How is the B200 different from the B300?** Both are 8-GPU Blackwell SXM nodes on the same terms. The B300 (Blackwell Ultra) carries 288 GB of memory per GPU to the B200’s 180 GB, so the B300 suits the largest models and the longest context windows. The B200 is the more cost-effective choice for large-model training and inference that fits comfortably in 180 GB per GPU. **What is the minimum order for B200 GPUs?** One node, which is 8 NVIDIA B200 SXM GPUs. We do not sell fewer than a full node, because the GPUs share NVLink and the node is the unit of allocation. Additional capacity is added in whole nodes. **What is the minimum commitment?** One year. This is dedicated bare-metal infrastructure reserved for you, not a spot instance. $3.75 per GPU-hour is our floor, on our longest term; a one-year term, the minimum, is priced higher. The term is billed upfront, with the finer details agreed case by case and confirmed in writing before you commit. **How much does a B200 node cost?** Pricing starts at $3.75 per GPU-hour, the floor rate on our longest commitments; it is higher on shorter terms and falls the longer you reserve. The minimum commitment is one year. The exact figure depends on your term length, region and the number of nodes, and we confirm it in writing before you commit. **What are the NVIDIA B200 specs?** The B200 is NVIDIA’s Blackwell generation. Each GPU has 180 GB of HBM3e and is built from two dies joined and presented to software as one device. It uses the second-generation Transformer Engine with 4-bit floating point (FP4) for inference, and fifth-generation NVLink at roughly 1.8 TB/s per GPU. An 8-GPU HGX B200 node holds 1,440 GB of HBM3e and comes air-cooled or liquid-cooled, depending on the server. **Is it HGX or DGX, and is it the same as the GB200?** Every B200 server we rent is an eight-GPU HGX B200 system: B200 SXM GPUs joined by NVLink at 1.8 TB/s each. The B200 has no PCIe version. NVIDIA’s own DGX B200 is built on the same board and adds NVIDIA’s software and support stack; tell us if you need DGX specifically. The GB200 is a different system: Blackwell GPUs paired with NVIDIA Grace CPUs in a 72-GPU NVL72 rack, which we do not rent. Our GB200 vs B200 explainer covers it. **Where are B200 nodes available?** India, the European Union and the United States. Blackwell capacity moves quickly, so we confirm B200 availability and lead time per region before you order. Pick the region by where your data has to live and where your team works. **What do you run, and what do I run?** We deliver the node as dedicated bare metal with network and root access, and keep the hardware running: continuous monitoring of GPU health and interconnect, hardware failure handling, and an engineer you can reach directly for the whole term. The software stack on top (OS, drivers, CUDA, container runtime and orchestration) is yours to run. **Can I scale beyond one node?** Yes. Nodes are added in units of eight B200 GPUs (1,440 GB of HBM3e each), and multi-node clusters are wired with a dedicated InfiniBand fabric, priced on request. Tell us the target size when we scope the workload and we will plan the interconnect and capacity for it. **Do you own the hardware?** No. We work with partners who own the hardware and hold the capacity in each region. FlexiCloud sources the node, delivers it to you as dedicated bare metal, and keeps the hardware running (monitored, failures handled and supported) for the whole term. The software you run on top is yours. ### More on running GPUs The maths and the trade-offs behind an 8-GPU node, cost, memory, rent vs buy, and the interconnect that makes it one machine. ### Tell us the workload and the region Fifteen minutes with an engineer gets you a firm price, a lead time, and a straight answer on whether the B200 or the B300 is the right fit. No commitment until you have the number in writing. ## NVIDIA B300 SXM GPU Cloud URL: https://flexicloud.co/gpu/nvidia-b300 What a B300 costs: an 8-GPU HGX B300 server sells for about US$490K to US$550K. Rent one in India, the EU or the US from $4.50 per GPU-hour. Last modified: 2026-10-26 ### Dedicated NVIDIA B300 SXM nodes, in India, the EU and the US Dedicated 8-GPU HGX B300 nodes on a one-year term. FlexiCloud delivers the bare metal and keeps the hardware running; you bring the workload and run your own stack. ### The numbers that matter ### Where the B300 earns its place The B300 is the headroom option: the most memory of any NVIDIA GPU we rent, for the models and runs that will not fit anywhere smaller. - Train the largest models with less sharding: 288 GB per GPU and 2,304 GB across the node means a model that would shard across many H100s can sit on far fewer B300s, which cuts communication overhead and simplifies the run. (Trillion-parameter and mixture-of-experts training; Long-context pretraining and post-training; Fewer GPUs per model, simpler scaling) - Serve very large models and long context: The second-generation Transformer Engine adds FP4, which roughly doubles inference throughput on suitable workloads, and 288 GB holds models that will not fit on smaller cards. (FP4 high-throughput inference; Long-context serving on a single GPU; Fewer replicas for the same model) - Choose the B300 when memory is the limit: Reach for the B300 when a model will not fit elsewhere without sharding, or when you want the most memory and the newest Blackwell Ultra performance. If 180 GB is enough, the B200 is the better value. (Most memory of any GPU we offer; Newest Blackwell Ultra generation; Weigh it against the B200 before you commit) ### The node is yours. The hardware is our job. Most GPU providers hand you a bare node and a support ticket queue. We keep the hardware running underneath: monitored around the clock, failures handled, and an engineer you can actually reach. - Full control of the stack: Root access to run your own OS image, CUDA version, container runtime and orchestration. No platform layer between you and the GPUs; the software is entirely yours. - Watched around the clock: GPU health, thermals, memory errors and interconnect are monitored continuously. A failing card is caught by us, not discovered by a crashed training run at 3am. - An engineer, not a queue: Support is a person who runs GPU infrastructure, in your timezone. When something is wrong you talk to someone who can fix it. - Data where you need it: Nodes in India for domestic data residency, and in the EU and US for teams and datasets that live there. Same service, same team, whichever region. - Yours alone: Dedicated bare-metal nodes. No shared tenancy, no noisy neighbours, no scheduler deciding when your job runs. The whole node is yours for the term. - Predictable cost: A fixed term at a fixed rate. No spot-market surprises, no burst pricing, no bill that doubles because a job ran long. You know the number for the whole year. ### How it is sold This is dedicated infrastructure on a term, not an hourly spot instance. The terms are simple and they are the same for everyone. - One node: 8 × B300 SXM: The unit is a full HGX B300 node. Eight GPUs with NVLink between them, dedicated to you. Need more? Nodes are added in units of eight. (8 × NVIDIA B300 SXM GPUs; 288 GB HBM3e per GPU, 2.3 TB per node; NVLink interconnect across the node; Dedicated bare metal, single tenant) - From $4.50 per GPU-hour: Our floor rate, on our longest term; the one-year minimum is priced higher. Billed upfront for the term, with the finer details agreed case by case. Final pricing depends on region, term length and how many nodes, and we confirm it in writing before you commit. (From $4.50 per GPU-hour; One-year minimum commitment; Longer terms and multi-node priced on request) - India, the EU and the US: Capacity is allocated per region on a first-committed basis. Tell us where the data lives and where the team sits, and we will confirm availability and lead time for that region before you commit to anything. (India: for domestic data residency; European Union: for EU-resident data and teams; United States: for US-based workloads; Lead time confirmed per region before order) ### From first conversation to a running node - Tell us the workload: Training, fine-tuning or inference; the framework; how many nodes; and which region. Fifteen minutes with an engineer, not a form that disappears into a CRM. - We confirm capacity and price: A firm quote for your region and term, with the lead time to a live node. No commitment until you have the number in writing. - We hand over the node: A dedicated bare-metal node, networked and ready to log into with root access. You bring your own OS image, drivers and orchestration; the hardware underneath stays ours to keep running. - You run. We keep it healthy.: Your team owns the OS, the stack and the workload. Ours watches the hardware, handles failures, and is a phone call away for the whole term. ### What people ask before they commit **What is the minimum order for B300 GPUs?** One node, which is 8 NVIDIA B300 SXM GPUs. We do not sell fewer than a full node, because the GPUs share NVLink and the node is the unit of allocation. Additional capacity is added in whole nodes. **What is the minimum commitment?** One year. This is dedicated bare-metal infrastructure reserved for you, not a spot instance. $4.50 per GPU-hour is our floor, on our longest term; a one-year term, the minimum, is priced higher. The term is billed upfront, with the finer details agreed case by case and confirmed in writing before you commit. **How much does it cost to buy an NVIDIA B300?** NVIDIA does not publish a price, and it sells B300 GPUs in 8-GPU HGX B300 servers rather than one at a time. Reuters reported in April 2026 that an 8-GPU B300 server cost about US$550,000 in the US, up from about US$500,000 in late 2025. A US GPU appraisal firm put a new HGX B300 server at US$490,000 to US$520,000 in October 2026 for a single unit, and US$400,000 to US$450,000 in volume. That is roughly US$50,000 to US$70,000 per GPU (our arithmetic), before power, cooling, networking and the data centre to run it. **What is the B300 price in India?** An 8-GPU B300 server at US$490,000 to US$550,000 is about ₹4.7 crore to ₹5.2 crore at ₹95.25 to the dollar, before import duty, GST and freight; buying also needs a DGFT import authorisation. Renting avoids that outlay: we source dedicated B300 nodes in India from $4.50 per GPU-hour on our longest term, about $26,280 a month for a whole node. **Is it cheaper to buy or rent a B300?** It depends on how long and how fully you will use it. Renting one node at our $4.50 floor rate costs about $315,360 a year, with power, cooling, hosting and hardware care included. Buying costs about half a million US dollars up front, plus a liquid-cooled data centre, power, networking and the people to run it, and the server loses value as NVIDIA's next generation, Vera Rubin, ships. Buying tends to win only with years of near-full use and your own data centre. Our rent, buy or API guide works through the numbers. **How much does it cost to rent a B300 node?** Pricing starts at $4.50 per GPU-hour, the floor rate on our longest commitments; it is higher on shorter terms and falls the longer you reserve. The minimum commitment is one year. The exact figure depends on your term length, region and the number of nodes, and we confirm it in writing before you commit. **What are the NVIDIA B300 specs?** The B300 is NVIDIA’s Blackwell Ultra generation. Each GPU has 288 GB of HBM3e, the most of any current NVIDIA GPU, and is built from two dies joined and presented to software as one device. It uses the second-generation Transformer Engine with 4-bit floating point (FP4) for inference, and fifth-generation NVLink at roughly 1.8 TB/s per GPU. An 8-GPU HGX B300 node holds 2,304 GB of HBM3e and is typically liquid-cooled. **Is it HGX or DGX, and is it the same as the GB300?** Every B300 server we rent is an eight-GPU HGX B300 system: B300 SXM GPUs with 288 GB each, joined by NVLink at 1.8 TB/s per GPU. NVIDIA announced the board as HGX B300 NVL16 because it counted each GPU’s two dies; your software sees eight GPUs. NVIDIA’s own DGX B300 is built on the same GPUs and adds NVIDIA’s software and support stack; tell us if you need DGX specifically. The GB300 is a different system: Blackwell Ultra GPUs paired with Grace CPUs in a 72-GPU NVL72 rack, which we do not rent. Our GB300 vs B300 explainer covers it. **Where are B300 nodes available?** India, the European Union and the United States. The B300 is the newest part we carry, so its allocation is the tightest of our GPUs: tell us your region early and we will confirm availability and lead time before you order. Pick the region by where your data has to live and where your team works. **What do you run, and what do I run?** We deliver the node as dedicated bare metal with network and root access, and keep the hardware running: continuous monitoring of GPU health and interconnect, hardware failure handling, and an engineer you can reach directly for the whole term. The software stack on top (OS, drivers, CUDA, container runtime and orchestration) is yours to run. **Can I scale beyond one node?** Yes. Nodes are added in units of eight B300 GPUs (2,304 GB of HBM3e each), and multi-node clusters are wired with a dedicated InfiniBand fabric, priced on request. Tell us the target size when we scope the workload and we will plan the interconnect and capacity for it. **Is this a shared or dedicated environment?** Dedicated. Each node is single-tenant bare metal reserved for you for the term. No shared GPUs, no scheduler queue, no noisy neighbours. **Do you own the hardware?** No. We work with partners who own the hardware and hold the capacity in each region. FlexiCloud sources the node, delivers it to you as dedicated bare metal, and keeps the hardware running (monitored, failures handled and supported) for the whole term. The software you run on top is yours. ### More on running GPUs The maths and the trade-offs behind an 8-GPU node, cost, memory, rent vs buy, and the interconnect that makes it one machine. ### Tell us the workload and the region Fifteen minutes with an engineer gets you a firm price, a lead time, and a straight answer on whether B300 is the right fit. No commitment until you have the number in writing. ## NVIDIA H100 SXM GPU Cloud URL: https://flexicloud.co/gpu/nvidia-h100 Rent dedicated 8-GPU NVIDIA H100 SXM servers in India, the EU and the US, from $2.95 per GPU-hour on the longest term. We keep the hardware running. Last modified: 2026-10-18 ### Dedicated NVIDIA H100 SXM nodes, in India, the EU and the US Dedicated 8-GPU HGX H100 nodes on a one-year term, the proven Hopper workhorse for training, fine-tuning and inference, at a lower rate than Blackwell. FlexiCloud delivers the bare metal and keeps the hardware running; you bring the workload and run your own stack. ### The numbers that matter ### Where the H100 earns its place The H100 is the proven Hopper workhorse: the widest software support and the lowest rate we offer, for the large share of workloads that fit in 80 GB. - The proven, cost-efficient workhorse: Two years of tuning mean most frameworks and kernels target the H100 first. For training and fine-tuning that fits in 80 GB, it is the lowest cost per useful result. (Training and fine-tuning that fits 80 GB; Widest framework and kernel support; Lowest GPU-hour rate we offer) - Reliable inference at the best rate: For models that fit in 80 GB, the H100 serves production inference at a lower rate than newer parts, with mature, well-understood performance and no surprises. (Steady production inference; Mature FP8 Transformer Engine; Predictable, well-documented behaviour) - Choose the H100 when the model fits and cost matters: Reach for the H100 when 80 GB is enough and you want proven, inexpensive Hopper compute. When a model or context will not fit in 80 GB, the H200 gives you more memory on the same generation. (80 GB HBM3 per GPU; The most available of our GPUs; Step up to the H200 for more memory) ### The node is yours. The hardware is our job. Most GPU providers hand you a bare node and a support ticket queue. We keep the hardware running underneath: monitored around the clock, failures handled, and an engineer you can actually reach. - Full control of the stack: Root access to run your own OS image, CUDA version, container runtime and orchestration. No platform layer between you and the GPUs; the software is entirely yours. - Watched around the clock: GPU health, thermals, memory errors and interconnect are monitored continuously. A failing card is caught by us, not discovered by a crashed training run at 3am. - An engineer, not a queue: Support is a person who runs GPU infrastructure, in your timezone. When something is wrong you talk to someone who can fix it. - Data where you need it: Nodes in India for domestic data residency, and in the EU and US for teams and datasets that live there. Same service, same team, whichever region. - Yours alone: Dedicated bare-metal nodes. No shared tenancy, no noisy neighbours, no scheduler deciding when your job runs. The whole node is yours for the term. - Predictable cost: A fixed term at a fixed rate. No spot-market surprises, no burst pricing, no bill that doubles because a job ran long. You know the number for the whole year. ### Get a quote for your workload Leave a name and a number or email. An engineer calls you back with pricing and availability for your region, usually within one working day. ### How it is sold This is dedicated infrastructure on a term, not an hourly spot instance. The terms are simple and they are the same for everyone. - One node: 8 × H100 SXM: The unit is a full HGX H100 node. Eight GPUs with NVLink between them, dedicated to you. Need more? Nodes are added in units of eight. (8 × NVIDIA H100 SXM GPUs; 80 GB HBM3 per GPU, 640 GB per node; NVLink interconnect across the node; Dedicated bare metal, single tenant) - From $2.95 per GPU-hour: Our floor rate, on our longest term; the one-year minimum is priced higher. Billed upfront for the term, with the finer details agreed case by case. Final pricing depends on region, term length and how many nodes, and we confirm it in writing before you commit. (From $2.95 per GPU-hour; One-year minimum commitment; Longer terms and multi-node priced on request) - India, the EU and the US: Capacity is allocated per region on a first-committed basis. Tell us where the data lives and where the team sits, and we will confirm availability and lead time for that region before you commit to anything. (India: for domestic data residency; European Union: for EU-resident data and teams; United States: for US-based workloads; Lead time confirmed per region before order) ### From first conversation to a running node - Tell us the workload: Training, fine-tuning or inference; the framework; how many nodes; and which region. Fifteen minutes with an engineer, not a form that disappears into a CRM. - We confirm capacity and price: A firm quote for your region and term, with the lead time to a live node. No commitment until you have the number in writing. - We hand over the node: A dedicated bare-metal node, networked and ready to log into with root access. You bring your own OS image, drivers and orchestration; the hardware underneath stays ours to keep running. - You run. We keep it healthy.: Your team owns the OS, the stack and the workload. Ours watches the hardware, handles failures, and is a phone call away for the whole term. ### What people ask before they commit **How is the H100 different from the B200 and B300?** The H100 is NVIDIA’s previous-generation Hopper GPU, with 80 GB of memory per GPU. The B200 and B300 are the newer Blackwell generation, with 180 GB and 288 GB per GPU and higher throughput. The H100 is the proven, cost-effective workhorse with the widest software and framework support; choose Blackwell when you need the extra memory for very large models and long context, or the newest performance. **What is the minimum order for H100 GPUs?** One node, which is 8 NVIDIA H100 SXM GPUs. We do not sell fewer than a full node, because the GPUs share NVLink and the node is the unit of allocation. Additional capacity is added in whole nodes. **What is the minimum commitment?** One year. This is dedicated bare-metal infrastructure reserved for you, not a spot instance. $2.95 per GPU-hour is our floor, on our longest term; a one-year term, the minimum, is priced higher. It is the lowest rate we offer. The term is billed upfront, with the finer details agreed case by case and confirmed in writing before you commit. **How much does an H100 node cost?** Pricing starts at $2.95 per GPU-hour, the floor rate on our longest commitments; it is higher on shorter terms and falls the longer you reserve. The minimum commitment is one year. The exact figure depends on your term length, region and the number of nodes, and we confirm it in writing before you commit. **What are the NVIDIA H100 specs?** The H100 is NVIDIA’s Hopper generation. Each GPU has 80 GB of HBM3 at roughly 3.4 TB/s. It uses the Transformer Engine with FP8, the workhorse precision for much of today’s training, and fourth-generation NVLink at roughly 900 GB/s per GPU. An 8-GPU HGX H100 node holds 640 GB of HBM3, and the ecosystem of drivers, kernels and frameworks around it is the most mature of any AI GPU. **Is it the SXM or PCIe H100, and is it HGX or DGX?** SXM, on an HGX board. Every H100 server we rent has eight H100 SXM GPUs with 80 GB each, joined through NVLink Switches at 900 GB/s per GPU, about seven times PCIe Gen 5. That is the same eight-GPU board NVIDIA uses in its own DGX H100; DGX adds NVIDIA’s software and support stack, so tell us if you need DGX specifically. The H100 also comes as a PCIe card, the H100 NVL with 94 GB, which we do not offer. Our guide to GPU form factors explains PCIe, SXM, HGX and DGX. **Where are H100 nodes available?** India, the European Union and the United States. Capacity is allocated per region, and we confirm it before you order. Pick the region by where your data has to live and where your team works. **What do you run, and what do I run?** We deliver the node as dedicated bare metal with network and root access, and keep the hardware running: continuous monitoring of GPU health and interconnect, hardware failure handling, and an engineer you can reach directly for the whole term. The software stack on top (OS, drivers, CUDA, container runtime and orchestration) is yours to run. **Can I scale beyond one node?** Yes. Nodes are added in units of eight H100 GPUs (640 GB of HBM3 each), and multi-node clusters are wired with a dedicated InfiniBand fabric, priced on request. Tell us the target size when we scope the workload and we will plan the interconnect and capacity for it. **Do you own the hardware?** No. We work with partners who own the hardware and hold the capacity in each region. FlexiCloud sources the node, delivers it to you as dedicated bare metal, and keeps the hardware running (monitored, failures handled and supported) for the whole term. The software you run on top is yours. ### Need more memory, or the newest generation? The H100 is the proven, lower-cost choice. Need more memory on the same generation? The H200 gives you 141 GB per GPU. For Blackwell-generation throughput and even more memory, the B200 and B300 step up from here. ### More on running GPUs The maths and the trade-offs behind an 8-GPU node, cost, memory, rent vs buy, and the interconnect that makes it one machine. ### Tell us the workload and the region Fifteen minutes with an engineer gets you a firm price, a lead time, and a straight answer on whether the H100, H200, B200 or B300 is the right fit. No commitment until you have the number in writing. ## NVIDIA H200 SXM GPU Cloud URL: https://flexicloud.co/gpu/nvidia-h200 Rent dedicated 8-GPU NVIDIA H200 SXM servers in India, the EU and the US. 141 GB HBM3e per GPU, from $3.35/GPU-hour on the longest term. Last modified: 2026-10-18 ### Dedicated NVIDIA H200 SXM nodes, in India, the EU and the US Dedicated 8-GPU HGX H200 nodes on a one-year term, the memory-upgraded Hopper with 141 GB of HBM3e per GPU, for inference, long context and larger models at a step below Blackwell. FlexiCloud delivers the bare metal and keeps the hardware running; you bring the workload and run your own stack. ### The numbers that matter ### Where the H200 earns its place The H200 is the memory-upgraded Hopper: the same compute as the H100 with nearly double the memory and bandwidth, for workloads the H100 runs out of room for. - Serve bigger models on fewer GPUs: 141 GB of HBM3e holds larger models and longer context than the H100’s 80 GB, so you serve them without sharding across extra cards, on the same CUDA stack you already run. (Large-model and long-context serving; Fewer GPUs per model than the H100; Same drivers and CUDA as the H100) - More headroom at Hopper cost: The H200 matches the H100 on compute but adds roughly 1.4x the bandwidth (around 4.8 TB/s) and nearly double the memory, which lifts throughput wherever the bottleneck is memory rather than maths. (Memory-bound fine-tuning; Higher inference throughput than the H100; About 4.8 TB/s of memory bandwidth) - Choose the H200 when the H100 runs out of memory: Reach for the H200 when your model or context will not fit in 80 GB but you do not need Blackwell’s price or FP4. If 80 GB is enough, the H100 is cheaper; if you need more still, the B200 steps up. (141 GB HBM3e per GPU; A step below Blackwell pricing; Same Hopper generation as the H100) ### The node is yours. The hardware is our job. Most GPU providers hand you a bare node and a support ticket queue. We keep the hardware running underneath: monitored around the clock, failures handled, and an engineer you can actually reach. - Full control of the stack: Root access to run your own OS image, CUDA version, container runtime and orchestration. No platform layer between you and the GPUs; the software is entirely yours. - Memory where it counts: 141 GB of HBM3e per GPU, nearly double the H100, at roughly 4.8 TB/s. Larger models and longer context fit without sharding, and memory-bound inference runs faster. - Watched around the clock: GPU health, thermals, memory errors and interconnect are monitored continuously. A failing card is caught by us, not discovered by a crashed training run at 3am. - An engineer, not a queue: Support is a person who runs GPU infrastructure, in your timezone. When something is wrong you talk to someone who can fix it. - Data where you need it: Nodes in India for domestic data residency, and in the EU and US for teams and datasets that live there. Same service, same team, whichever region. - Predictable cost: A fixed term at a fixed rate. No spot-market surprises, no burst pricing, no bill that doubles because a job ran long. You know the number for the whole year. ### Get a quote for your workload Leave a name and a number or email. An engineer calls you back with pricing and availability for your region, usually within one working day. ### How it is sold This is dedicated infrastructure on a term, not an hourly spot instance. The terms are simple and they are the same for everyone. - One node: 8 × H200 SXM: The unit is a full HGX H200 node. Eight GPUs with NVLink between them, dedicated to you. Need more? Nodes are added in units of eight. (8 × NVIDIA H200 SXM GPUs; 141 GB HBM3e per GPU, 1,128 GB per node; NVLink interconnect across the node; Dedicated bare metal, single tenant) - From $3.35 per GPU-hour: Our floor rate, on our longest term; the one-year minimum is priced higher. Billed upfront for the term, with the finer details agreed case by case. Final pricing depends on region, term length and how many nodes, and we confirm it in writing before you commit. (From $3.35 per GPU-hour; One-year minimum commitment; Longer terms and multi-node priced on request) - India, the EU and the US: Capacity is allocated per region on a first-committed basis. Tell us where the data lives and where the team sits, and we will confirm availability and lead time for that region before you commit to anything. (India: for domestic data residency; European Union: for EU-resident data and teams; United States: for US-based workloads; Lead time confirmed per region before order) ### From first conversation to a running node - Tell us the workload: Training, fine-tuning or inference; the framework; how many nodes; and which region. Fifteen minutes with an engineer, not a form that disappears into a CRM. - We confirm capacity and price: A firm quote for your region and term, with the lead time to a live node. No commitment until you have the number in writing. - We hand over the node: A dedicated bare-metal node, networked and ready to log into with root access. You bring your own OS image, drivers and orchestration; the hardware underneath stays ours to keep running. - You run. We keep it healthy.: Your team owns the OS, the stack and the workload. Ours watches the hardware, handles failures, and is a phone call away for the whole term. ### What people ask before they commit **How is the H200 different from the H100?** The H200 is the same Hopper generation and the same compute as the H100, with one big change: memory. It has 141 GB of HBM3e per GPU at roughly 4.8 TB/s, against the H100’s 80 GB of HBM3 at around 3.4 TB/s. That nearly doubles the room per GPU, so larger models and longer context fit without sharding, and memory-bound inference runs noticeably faster. If your workload is limited by memory rather than raw compute, the H200 is the upgrade, for a small step up in price. **How is the H200 different from the B200 and B300?** The H200 is Hopper; the B200 and B300 are the newer Blackwell generation, with 180 GB and 288 GB per GPU, a 4-bit inference format and roughly double the NVLink bandwidth. The H200 is the value step when you need more memory than an H100 gives you but not Blackwell’s price or performance. Choose Blackwell for the very largest models, the heaviest training, or when FP4 inference changes your economics. **What is the minimum order for H200 GPUs?** One node, which is 8 NVIDIA H200 SXM GPUs. We do not sell fewer than a full node, because the GPUs share NVLink and the node is the unit of allocation. Additional capacity is added in whole nodes. **What is the minimum commitment?** One year. This is dedicated bare-metal infrastructure reserved for you, not a spot instance. $3.35 per GPU-hour is our floor, on our longest term; a one-year term, the minimum, is priced higher. The term is billed upfront, with the finer details agreed case by case and confirmed in writing before you commit. **How much does an H200 node cost?** Pricing starts at $3.35 per GPU-hour, the floor rate on our longest commitments; it is higher on shorter terms and falls the longer you reserve. The minimum commitment is one year. The exact figure depends on your term length, region and the number of nodes, and we confirm it in writing before you commit. **What are the NVIDIA H200 specs?** The H200 is NVIDIA’s Hopper generation, refreshed. It has the same compute as the H100 but 141 GB of HBM3e per GPU at roughly 4.8 TB/s, against the H100’s 80 GB at around 3.4 TB/s. It uses the Transformer Engine with FP8, and fourth-generation NVLink at roughly 900 GB/s per GPU. An 8-GPU HGX H200 node holds 1,128 GB of HBM3e. **Is it the SXM or NVL H200, and is it HGX or DGX?** SXM, on an HGX board. Every H200 server we rent has eight H200 SXM GPUs with 141 GB each, joined through NVLink Switches at 900 GB/s per GPU. That is the same eight-GPU board NVIDIA uses in its own DGX H200; tell us if you need DGX specifically. The H200 NVL is a different product: a PCIe card that bridges two or four cards over NVLink, without the all-to-all NVLink Switch fabric. We do not offer it. Our guide to GPU form factors explains the difference. **Where are H200 nodes available?** India, the European Union and the United States. The H200 is in steadier supply than the Blackwell parts, but capacity is still allocated per region, so we confirm H200 availability and lead time for your region before you order. Pick the region by where your data has to live and where your team works. **What do you run, and what do I run?** We deliver the node as dedicated bare metal with network and root access, and keep the hardware running: continuous monitoring of GPU health and interconnect, hardware failure handling, and an engineer you can reach directly for the whole term. The software stack on top (OS, drivers, CUDA, container runtime and orchestration) is yours to run. **Can I scale beyond one node?** Yes. Nodes are added in units of eight H200 GPUs (1,128 GB of HBM3e each), and multi-node clusters are wired with a dedicated InfiniBand fabric, priced on request. Tell us the target size when we scope the workload and we will plan the interconnect and capacity for it. **Do you own the hardware?** No. We work with partners who own the hardware and hold the capacity in each region. FlexiCloud sources the node, delivers it to you as dedicated bare metal, and keeps the hardware running (monitored, failures handled and supported) for the whole term. The software you run on top is yours. ### Need it cheaper, or need Blackwell? The H200 is the memory-upgraded Hopper. If cost matters more than memory, the H100 is lower; if you need more memory still or Blackwell-generation throughput, the B200 and B300 step up from here. ### More on running GPUs The maths and the trade-offs behind an 8-GPU node, cost, memory, rent vs buy, and the interconnect that makes it one machine. ### Tell us the workload and the region Fifteen minutes with an engineer gets you a firm price, a lead time, and a straight answer on whether the H100, H200, B200 or B300 is the right fit. No commitment until you have the number in writing. ## NVIDIA RTX PRO 6000 GPU Cloud URL: https://flexicloud.co/gpu/nvidia-rtx-pro-6000 NVIDIA RTX PRO 6000 Blackwell: 96 GB GDDR7, full specs, how it compares with the RTX 6000 Ada, what a card costs, and cloud access. Register interest. Last modified: 2026-10-16 ### NVIDIA RTX PRO 6000 Blackwell is coming to FlexiCloud NVIDIA’s data-centre RTX GPU: 96 GB of GDDR7 per card, for inference, video, rendering and virtual workstations. Register now and we will come to you first with pricing and availability. ### The card, in numbers ### One GPU for AI, graphics and video NVIDIA positions the RTX PRO 6000 as a universal data-centre GPU: Blackwell Tensor Cores for AI, plus RT Cores for ray tracing and dedicated video engines. These are the workloads it suits best. - Inference for small and mid-size models: 96 GB per GPU holds a 70B-parameter model at FP8 on a single card with room for modest context, and FP4 support stretches that further. Run one model per GPU, or eight separate services on one server. - Multimodal, vision and video: Four NVENC encoders, four NVDEC decoders and four JPEG engines per GPU sit alongside the Tensor Cores. That suits video search and summarisation, vision models, speech, and transcoding at volume. - Rendering and graphics: 188 fourth-generation RT Cores bring full RTX ray tracing to the data centre: final-frame and batch rendering and neural rendering, without tying up a desk-side workstation. - Simulation and digital twins: Strong FP32 throughput and RTX graphics make it a fit for Omniverse and OpenUSD pipelines, factory and robotics simulation, and synthetic data generation for physical AI. - Virtual workstations and shared GPUs: Multi-Instance GPU splits each card into up to four isolated 24 GB instances, and the card supports NVIDIA’s virtual workstation software for VDI, licensed separately by NVIDIA. Several teams or users can share one server safely. - Same model as every FlexiCloud GPU: We intend to offer it the way we offer every GPU: dedicated bare metal with root access, sourced across our partner network in India, the EU and the US. We keep the hardware running; you run your stack. ### Register early interest in the RTX PRO 6000 Leave a name and a number or email. We will come to you first with pricing and availability, and if you need compute sooner we can put you on an H100 or H200 node now. ### Honest status, so you can plan The card is not new; our offer is. Here is what is confirmed, where it fits and where it does not, and what registering gets you. - Shipping now in NVIDIA RTX PRO Servers: The RTX PRO 6000 Blackwell Server Edition has been in production since 2025, in servers with two, four or eight GPUs. What we are still finalising is the FlexiCloud offer: pricing, terms and regions across our partner network. (96 GB GDDR7 per GPU at 1.6 TB/s; PCIe Gen 5, no NVLink between GPUs; Up to four isolated MIG instances per GPU) - Training at scale belongs on SXM nodes: Without NVLink, the GPUs in an RTX PRO server talk over PCIe and cannot pool memory at speed, so eight of them behave more like eight separate GPUs than one machine. For pre-training and large fine-tunes, our B300, B200, H200 and H100 nodes are the right tool, and they are available now. (Pre-training and multi-node training; Large fine-tunes spread across many GPUs; Models that must be split across GPUs at speed) - Register now, no commitment: Registering costs you nothing and commits you to nothing. It puts you on our early list for the first quote, and we take your requirement across our partner network in confidence, not to ten sales teams. (First to get pricing and availability; One quote across 10+ partner providers; No markup from us) ### What people are asking about the RTX PRO 6000 **What is the NVIDIA RTX PRO 6000 Blackwell Server Edition?** It is the data-centre version of NVIDIA’s RTX PRO 6000 Blackwell: a passively cooled, dual-slot PCIe Gen 5 card with 96 GB of GDDR7 memory, fifth-generation Tensor Cores with FP4, fourth-generation RT Cores and dedicated video engines. NVIDIA positions it as a universal data-centre GPU for AI inference, graphics, simulation and video, and it ships in servers from the major manufacturers with two, four or eight GPUs. **What is the RTX PRO 6000 good for, and what is it not for?** It suits inference of small and mid-size models (a 70B-parameter model fits on one card at FP8), multimodal and video AI, rendering, simulation and digital twins, and virtual workstations. Light fine-tuning of smaller models on a single GPU is fine. It is not the right choice for pre-training or large fine-tunes across many GPUs: it has no NVLink, so the GPUs talk over PCIe and cannot pool memory at speed. For that work, our B300, B200, H200 and H100 nodes are available now. **What are the RTX PRO 6000 Blackwell Server Edition specs?** Per GPU, from NVIDIA’s published specifications: 96 GB of GDDR7 with ECC on a 512-bit bus, 1,597 GB/s of memory bandwidth, 24,064 CUDA cores, 752 fifth-generation Tensor Cores and 188 fourth-generation RT Cores. NVIDIA’s peak figures are up to 4 PFLOPS at FP4 and 2 PFLOPS at FP8 (both with sparsity), and about 120 TFLOPS at FP32. It draws up to 600 W (configurable), connects over PCIe Gen 5 x16, has four NVENC and four NVDEC engines, supports up to four MIG instances of 24 GB each, and has no NVLink. **How much does an RTX PRO 6000 cost?** To buy one card outright: NVIDIA’s US store listed the RTX PRO 6000 Blackwell Workstation Edition at US$16,000 in August 2026, up from US$13,250 in June (Tom’s Hardware). The Server Edition is sold through server makers, usually inside 8-GPU servers, and is priced on quote. Renting avoids that outlay and the power, cooling and hosting that come with it. Our per-GPU-hour pricing for the RTX PRO 6000 is to be announced; register interest and you will be among the first to get a firm quote. **How does the RTX PRO 6000 compare with the RTX 6000 Ada?** The RTX PRO 6000 Blackwell is the successor to the RTX 6000 Ada Generation. It doubles the memory to 96 GB of GDDR7 (the Ada card has 48 GB of GDDR6) and raises memory bandwidth from 960 GB/s to 1,597 GB/s. It has 24,064 CUDA cores against 18,176, about 120 TFLOPS of FP32 against 91.1, adds FP4 support and up to four MIG instances, and moves to PCIe Gen 5. The Ada card draws 300 W; the Server Edition draws up to 600 W. Figures are NVIDIA’s published specifications. **How is the RTX PRO 6000 different from the H100 and H200?** The H100 and H200 are SXM parts with HBM memory and NVLink, so eight GPUs work together as one machine for training and for large models. The RTX PRO 6000 is a PCIe card with GDDR7: more memory per GPU than the H100 (96 GB against 80 GB) but lower memory bandwidth, no NVLink, and RT Cores and video encoders that the H100 and H200 do not have. Choose the H100 or H200 for training and for models split across GPUs; choose the RTX PRO 6000 for inference that fits on one card, graphics, video and simulation. Our RTX PRO 6000 vs H100 guide, linked below, compares the two in plain language. **When will the RTX PRO 6000 be available on FlexiCloud, and what will it cost?** The card itself is shipping now. We are finalising pricing, terms and regions with our partner providers, so pricing is to be announced and we will not quote a number until it is firm. Register your interest and you will be among the first to get a firm quote. If you need compute sooner, we can put you on an H100 or H200 node now. **Can one RTX PRO 6000 be shared between users or workloads?** Yes. Multi-Instance GPU (MIG) splits each card into up to four fully isolated instances of 24 GB, each with its own memory, cache and compute, so an eight-GPU server can run up to 32 isolated workloads. The card also supports NVIDIA’s virtual workstation and virtual PC software (vWS and vPC) for VDI, which NVIDIA licenses separately. **Why register through FlexiCloud rather than going direct?** FlexiCloud sources GPU capacity across a network of more than ten partner providers in India, the EU and the US. You get one relationship and one quote across all of them, with no markup from us, because we are paid by the providers, not by you. Your requirement stays in confidence instead of going to ten sales teams, and you are not tied to one data centre or region. ### Before you choose a GPU How it compares with the H100, whether your model fits in 96 GB, why NVLink matters for training, and how buying through us works. ### Get on the RTX PRO 6000 early list Tell us the workload, how many GPUs and the region, and we will come to you first with pricing and availability. Need compute today? Our B300, B200, H200 and H100 nodes are available now. ## NVIDIA Vera Rubin GPU Cloud URL: https://flexicloud.co/gpu/nvidia-vera-rubin NVIDIA Vera Rubin, 288 GB of HBM4 per GPU, is in full production; we expect wider availability in 2027. Register early interest in Rubin nodes. Last modified: 2026-10-24 ### NVIDIA Vera Rubin is coming to FlexiCloud NVIDIA’s generation after Blackwell: the Rubin GPU with 288 GB of HBM4, paired with the new Vera CPU. Register now and we will come to you first with pricing and availability. ### Vera Rubin at a glance ### The generation after Blackwell Based on NVIDIA’s published specifications for Vera Rubin, which may still be revised. Where you need compute today, the Blackwell B200 and B300 are available now. - HBM4 memory: The Rubin GPU moves to HBM4: 288 GB per GPU and up to 19.2 TB/s of memory bandwidth on NVIDIA’s figures, more than twice the B300’s 8 TB/s of HBM3e. More room and more bandwidth for the biggest models and longest context. - A big FP4 inference leap: NVIDIA rates each Rubin GPU at up to 50 petaflops of NVFP4 inference, a large step up from Blackwell. The target is lower cost per token at frontier scale, and NVIDIA’s roadmap places Rubin Ultra in 2027. - Vera CPU and NVLink 6: Rubin pairs with the new Vera CPU and sixth-generation NVLink. The rack-scale Vera Rubin NVL72 wires 72 Rubin GPUs and 36 Vera CPUs together as one accelerator, and the 8-GPU HGX Rubin NVL8 brings the same GPU to a standard server node. - Built on a 3nm-class process (reported): Rubin is reported to be built on a 3nm-class process, a step on from Blackwell. At rack scale it is liquid-cooled, like the top Blackwell systems. - Same regions, same model: We intend to offer Rubin the way we offer every GPU: dedicated bare metal with root access, in India, the EU and the US, with the hardware kept running by us and the stack run by you. - Reserved, not speculative: Rubin capacity is scarce, and the first racks are going to large cloud and AI-infrastructure partners. Registering interest now puts you in the queue for a firm quote, and for allocation, as soon as pricing is set. ### Register early interest in Vera Rubin Leave a name and a number or email, and tell us roughly what you are planning: how many GPUs, the workload, and the region. We will come to you first with Rubin pricing and availability, and in the meantime we can get you onto Blackwell if the timeline is tight. ### Honest status, so you can plan Rubin is new. Here is what is known, when to expect it, and what registering gets you, with nothing oversold. - In full production: NVIDIA says Vera Rubin is in full production, and its August 2026 results reported racks already running at cloud and AI-infrastructure partners. The figures on this page come from NVIDIA’s published specifications and may be revised. (Rubin GPU with 288 GB of HBM4; Paired with the new Vera CPU; NVLink 6, the NVL72 rack and the 8-GPU HGX Rubin NVL8) - We expect wider availability in 2027: The first racks are running at large cloud and AI-infrastructure partners now. We expect wider availability through our partner network in 2027. That is our estimate, not an NVIDIA date: exact timing depends on NVIDIA’s allocation, and we confirm it in writing once it is firm. (Rubin: shipping now, to large partners first; Wider availability: 2027 (our estimate); Rubin Ultra: 2027, on NVIDIA’s roadmap) - Register now, no commitment: Registering interest costs you nothing and commits you to nothing. It puts you on our early list, so you get the first quote and a shot at an allocation slot before general availability. (First to get pricing and availability; Early access to launch allocation where we can secure it; We can bridge you onto Blackwell meanwhile) ### What people are asking about Rubin **What is NVIDIA Vera Rubin?** Vera Rubin is NVIDIA’s GPU platform after Blackwell, named after the astronomer Vera Rubin. It pairs a new GPU called Rubin, with 288 GB of HBM4 memory, with a new CPU called Vera. At rack scale it is the Vera Rubin NVL72, which connects 72 Rubin GPUs and 36 Vera CPUs over NVLink 6 as a single accelerator. NVIDIA’s earlier roadmap called the same rack NVL144, counting GPU dies rather than GPU packages. There is also an 8-GPU server form, the HGX Rubin NVL8. **When will Vera Rubin be available?** NVIDIA says Vera Rubin is in full production, and its first racks are running at large cloud and AI-infrastructure partners. We expect wider availability through our partner network in 2027. That is our estimate, not an NVIDIA date, and it depends on NVIDIA’s allocation. Rubin Ultra follows on NVIDIA’s roadmap in 2027. Register your interest and we will tell you the moment we have firm FlexiCloud availability. **How much will a Rubin node cost on FlexiCloud?** Pricing is to be announced. We will quote a rate once allocation and pricing are confirmed with our partners. Register your interest and you will be among the first to get a firm number, before general availability. **How is Rubin different from the B200 and B300?** The B200 and B300 are the current Blackwell generation, available now with up to 288 GB of HBM3e. Rubin is the next generation: 288 GB of HBM4 at more than twice the B300’s memory bandwidth, a large jump in FP4 inference performance on NVIDIA’s figures, and NVLink 6. If you need compute today, Blackwell is the answer; Rubin is for planning a 2027 buildout. **What is the difference between Rubin and Rubin Ultra?** Rubin is the first part of the generation, and NVIDIA says production shipments began in August 2026. Rubin Ultra is the higher-end follow-on, which NVIDIA’s March 2026 roadmap places in 2027 with more memory per GPU. Analysts have since reported design changes and a slip to 2028 for its largest rack; NVIDIA says its roadmap is intact. We intend to offer both as they become available. **How do I reserve Rubin capacity?** Register your interest through the form on this page with a rough idea of how many GPUs, the workload and the region. There is no cost and no commitment. It puts you on our early list for the first quote and early access to launch allocation where we can secure it, and we can put you on Blackwell in the meantime if your timeline is tight. ### While you wait for Rubin What Rubin changes and when it matters, how the current generations compare, and how to size memory, so the Blackwell decision you make today still fits when Rubin lands. ### Get on the Rubin early list Tell us the rough shape of what you are planning and we will come to you first with pricing and availability. Need compute sooner? We will get you onto a Blackwell node now. ## Dedicated NVIDIA GPU nodes in India, the EU and the US URL: https://flexicloud.co Dedicated 8-GPU NVIDIA B300, B200, H200 and H100 nodes in India, the EU and the US, from $2.95 per GPU-hour. One quote, with no markup from us. Last modified: 2026-10-18 ### Dedicated NVIDIA GPUs, from H100 to B300 Dedicated 8-GPU nodes for training and inference, sourced across more than ten partner providers. One quote, with no markup from us; you run the stack, we keep the hardware running. ### NVIDIA GPU lineup - NVIDIA B300: from $4.50 per GPU-hour; 288 GB HBM3e per GPU; available now. Our largest-memory GPU available today, for the biggest models and the longest context. - NVIDIA B200: from $3.75 per GPU-hour; 180 GB HBM3e per GPU; available now. Cost-effective Blackwell for large-model training and FP4 inference. - NVIDIA H200: from $3.35 per GPU-hour; 141 GB HBM3e per GPU; available now. Hopper with more memory, for long context and models that outgrow 80 GB. - NVIDIA H100: from $2.95 per GPU-hour; 80 GB HBM3 per GPU; available now. The proven workhorse, at our lowest rate. - NVIDIA Vera Rubin: pricing to be announced; 288 GB HBM4 per GPU; coming soon. The generation after Blackwell. In full production at NVIDIA; we expect wider availability in 2027. - NVIDIA RTX PRO 6000: pricing to be announced; 96 GB GDDR7 per GPU; coming soon. Inference, video, rendering and virtual workstations on NVIDIA’s data-centre RTX GPU. Floor rates in USD on our longest term; shorter terms cost more. Minimum one 8-GPU node on a one-year commitment. Your rate depends on region, term and node count, and is confirmed in writing before you commit. ### One relationship. More than ten providers behind it. We are not another GPU provider. We are the independent layer above them: we find the right node across our partner network, deliver it as dedicated bare metal, and keep the hardware running for the term. - One quote across the network: Describe the workload once. We check it against capacity across our partners in India, the EU and the US, and come back with real options, not ten sales calls. - No markup from us: Providers pay our commission out of their own sales budget, so nothing is added to your rate. And because we place volume across the network, we know where the real floor is. - Priority on scarce parts: A request through us carries our volume and standing with the provider. When B300 and B200 allocation is tight, that moves you up the queue. - Live inventory, not one data centre: We see what partners actually have this week, under NDA, region by region. If one is full, another may have exactly what you need. Capacity we cannot verify is never offered. - Your plans stay private: We scope your requirement in confidence and approach only the providers likely to fit it. Your roadmap is not broadcast to ten sales teams. - One contact, little to no upselling: One person whose job is the right node at the right price, and an engineer on the hardware for the whole term. We have no platform of our own to push. ### How GPU nodes are sold Dedicated infrastructure on a term, not hourly spot instances. The terms are simple, and they are the same for every GPU available today. - One node of 8 GPUs: The unit is a full 8-GPU SXM node with NVLink between the GPUs, single tenant and dedicated to you. Need more? Add whole nodes; multi-node clusters get a dedicated InfiniBand fabric, priced on request. (Bare metal with root access; Your OS, drivers, CUDA and orchestration; Hardware monitored, failures handled by us) - One-year minimum, fixed rate: The rate shown for each GPU is the floor, on our longest term; shorter terms cost more. Your rate depends on region, term and node count, and is confirmed in writing before you commit. No burst pricing, no surprise bills. (One-year minimum commitment; Billed upfront for the term; Quote in writing before you sign) - India, the EU and the US: Capacity is allocated per region on a first-committed basis. Tell us where the data has to live and we confirm availability and the lead time to a live node before you sign anything. (India: for domestic data residency; European Union: for EU-resident data and teams; United States: for US-based workloads; Lead time in writing, per region) ### Get GPU pricing and availability from across the network Leave a name and a number or email. We call you back with real options for your GPU, region and term, usually within one working day. ### What GPU buyers ask before they commit **Which NVIDIA GPUs can I rent from FlexiCloud today?** Four, each as a dedicated 8-GPU SXM node in India, the EU or the US: the B300 with 288 GB of HBM3e per GPU, the B200 with 180 GB of HBM3e, the H200 with 141 GB of HBM3e and the H100 with 80 GB of HBM3. NVIDIA Vera Rubin and the RTX PRO 6000 Blackwell Server Edition are coming soon, and you can register interest in either now. **How much does it cost to rent a GPU node?** Rates start at $2.95 per GPU-hour for the H100, $3.35 for the H200, $3.75 for the B200 and $4.50 for the B300. Those are floor rates in USD on our longest term; shorter terms cost more. Your exact rate depends on the GPU, region, term and number of nodes, and we confirm it in writing before you commit. **Can I rent a single GPU, or pay by the hour?** No. The unit is a full node of 8 GPUs, because the GPUs share NVLink and are allocated together, and the minimum term is one year. Prices are quoted per GPU-hour so they are easy to compare, but this is dedicated bare metal reserved for you, not an on-demand or spot instance. More capacity is added in whole nodes, and each node is billed upfront for its term. **How quickly can I get a node?** It depends on the GPU and the region, so we do not publish a lead time we cannot stand behind. Because we see live capacity across our partner providers, we can usually tell you within one working day what is available and how soon, and we put the lead time in writing before you commit. B300 allocation is the tightest, so ask early. **Why rent GPUs through FlexiCloud rather than going direct to a provider?** One relationship instead of ten sales conversations. We work with more than ten GPU providers in India, the EU and the US, so one request is checked against live inventory across all of them; our volume gets you priority on scarce parts; you are not tied to one data centre or region; and your requirement is not broadcast to ten sales teams. Providers pay our commission out of their own sales budget, not as a markup on your rate, so we add nothing to your rate. **Do you own any of the GPUs?** Not today. Every node we offer runs on our partners’ hardware, which is why we can quote across more than ten providers. We are working with partners towards deploying some FlexiCloud-owned capacity in a few locations. When that happens, it will be quoted alongside partner options in exactly the same way, and you will always know whose hardware you are on. **What do you manage, and what do I manage?** FlexiCloud sources the node, delivers it to you as dedicated, single-tenant bare metal with root access, and keeps the hardware running for the term: GPU health and interconnect monitored, failures handled, and an engineer you can reach. The OS, drivers, CUDA, containers and orchestration are yours to run. **When will NVIDIA Vera Rubin and the RTX PRO 6000 be available?** Both are coming soon to FlexiCloud, with pricing to be announced. NVIDIA says Vera Rubin is in full production, with its first racks running at large cloud and AI-infrastructure partners; we expect wider availability through our partner network in 2027. The RTX PRO 6000 Blackwell Server Edition, with 96 GB of GDDR7 per GPU, already ships from NVIDIA, and we are finalising our offer for it. Register interest on either page and we will come to you first. If you need compute now, the B300, B200, H200 and H100 are available today. **Do you still offer managed cloud hosting?** Yes. Managed cloud hosting on DigitalOcean, Vultr, Hetzner and Linode, shared hosting and the FlexiCloud Panel carry on as before, with free migration and the same team. You will find it all on our managed cloud hosting page. ### Offering GPU capacity? We act as an extended sales team for GPU owners: we qualify buyers and bring them to you ready to talk, with your details shared only under NDA. No listing fees; commission only when a deal closes. ### Choosing a GPU, and how to buy one How the broker model works, how the GPU generations compare, whether your model fits, and whether to rent, buy or use an API at all. ### Tell us the GPU, how many nodes and the region We come back with real options from across our network, usually within one working day, with the price and lead time in writing before you commit to anything. ## Managed cloud hosting on the provider you choose URL: https://flexicloud.co/managed-cloud Managed hosting on DigitalOcean, Vultr, Hetzner or Linode, run 24×7 by engineers in your timezone. From $10/month, free migration, no long-term contract. Last modified: 2026-10-18 ### Your cloud. Our operations team. Enterprise-grade hosting on DigitalOcean, Vultr, Hetzner or Linode. Provisioned, tuned and watched 24×7 by engineers in your timezone. ### Start where you are. Move up when you outgrow it. Most of our cloud customers started on shared hosting. When a site earns more than it costs to run properly, we move it, free and without downtime. - Get online, cheaply: Your site on a well-maintained shared server with the Webuzo control panel, free SSL and email. The right choice for brochure sites, portfolios and projects finding their feet. (Webuzo panel and one-click installs; Free SSL and nightly backups; Unmetered bandwidth and email; From ₹500 per month) - Your own server, run for you: Dedicated CPU and memory on the cloud provider you choose, with the full stack built, secured and monitored by our team. For sites where slow costs money. (Dedicated resources: no noisy neighbours; One-click restores and atomic deploys; Scale the server up or down as traffic moves; From $10 per month) ### The work that happens after the server exists Anyone can hand you a cloud server in sixty seconds. Keeping it fast, patched and online for three years is the job you are actually paying for. - Provisioned and tuned: We size the server, build the stack, and tune LiteSpeed, PHP, Redis and MySQL for your workload. You get a running site, not an empty Ubuntu box. - Patched and hardened: OS and stack updates, firewall rules, SSL renewal and malware scanning are ours to keep current, including the ones that land at 2am. - Watched around the clock: Uptime, CPU, memory and disk are monitored continuously. When something moves the wrong way, an engineer is already looking at it. - Backups you can restore: Offsite backups with one-click restore, with storage billed at $0.033 per GB per month. We test restores, because a backup you have never restored is not a backup. - Scaled without downtime: Traffic spike or a sale coming up? We resize the server or add capacity, usually while the site stays up. - Supported by people: Real engineers on chat, ticket and phone, in your timezone, not eleven hours behind it. ### One panel for every server you run Deploy a site, restore a backup, invite a client, read your invoice, without a support ticket and without touching SSH. ### Moving your site is our job, not your weekend Free on every plan, scheduled around you, with the old host kept as a fallback. - Tell us what you run: Send us the current host, the stack and a rough traffic figure. No access needed yet. - We size and build: We recommend a provider and plan, then build the server and copy the site across. - You test on a temp URL: The site runs on a temporary address until you have clicked through it and signed off. - We switch DNS: Cutover happens at a time you choose, with the old host kept warm as a fallback. ### Before you move your site **Who owns the cloud server, you or me?** You choose the provider and the plan; we provision and operate it. If you ever leave, your site and data come with you. There is no proprietary format to unpick. **What is the difference between your shared hosting and managed cloud?** Shared hosting puts your site on a server alongside others: the cheapest way to be online, and fine for brochure sites and early projects. Managed cloud gives you your own server with dedicated CPU and memory, so traffic on someone else’s site can never slow yours down. **Can you move my existing site?** Yes, and migration is free. We copy the site to the new server, test it on a temporary URL, and only switch DNS once you have signed off. **Which cloud provider should I pick?** For most Indian traffic, DigitalOcean Bangalore or Vultr Mumbai. For European audiences, Hetzner is the best value per core. If you are unsure, tell us where your visitors are and we will recommend one. **What happens if my site goes down at 3am?** Our monitoring pages an on-call engineer. You do not have to notice the problem first, and you do not have to wait for a US business day. **Is there a contract or a refund?** No long-term contract; plans are monthly or yearly and you can cancel any time. Shared hosting and monthly managed cloud plans carry a 30-day money-back guarantee for first-time customers; GPU nodes are sold on a committed term, billed upfront, and are not covered. ### Worth reading before you choose ### Tell us what you are running today. We will tell you what it should be running on, and move it there for free. ## Partner with FlexiCloud URL: https://flexicloud.co/partners Own or run GPU capacity? FlexiCloud brings qualified buyers to H100, H200, B200 and B300 providers. Commission only on closed deals. Details stay private. Last modified: 2026-10-14 ### Got GPU capacity? We bring the buyers. FlexiCloud works with more than ten GPU providers. We act as your extended sales team: we find serious buyers, qualify what they need, and bring them to you ready to talk. You pay a commission only when a deal closes. ### Customers, without building a sales team You run the hardware. We do the part that is slow and expensive for a provider: finding buyers who are ready to commit. - Buyers who are ready: We qualify the requirement, term, region and budget before anyone is introduced to you, so the conversations you have are worth having. - Little effort on your side: One point of contact, and no sales team needed. We handle the enquiry, the back and forth and the follow-up, then bring you the deal. - Pay only on results: There is no fee to be on our books. We earn a commission when a deal closes, so we only win when you do. - Your details stay private: We never publish your inventory, prices or name. A buyer sees your details only under NDA, and only with your go-ahead. - More than one door: Your capacity is matched against demand from India, the EU and the US, so it is not limited to the buyers one sales team happens to know. - You talk to the founder: You deal with Vinod Chacko directly. 21 years in infrastructure, cloud and AI. ### From your offer to a signed deal - Tell us what you have: Fill in the form below. It takes about three minutes: the GPU model, how many, where it is, who controls it, and your terms. - We sign an NDA: Before either side shares details, we sign a mutual NDA, so your inventory and pricing stay confidential. - We verify it: A short call, then checks on ownership, location, power, networking and SLA. We ask for references. Capacity we cannot verify is never offered to a buyer. - We match it to buyers: We put your verified capacity in front of the buyers whose requirement fits, and we tell you when there is real interest. - You close, we are paid: The deal closes on terms agreed with you. Our commission is due only when it does. ### What we are looking for We would rather say no early than waste your time, so here is the shape of capacity we can place. - Dedicated GPU nodes: Bare-metal or dedicated nodes we can offer to buyers on a term. (NVIDIA H100, H200, B200 and B300, and Vera Rubin when it ships; Full 8-GPU nodes, one node at minimum; India first; the EU and the US are welcome; Terms of 12 months or more suit most buyers) - The details that decide a deal: The more precise you are, the faster we can tell you if we have a buyer. (Who owns or controls the hardware; Exact location, power and cooling; Networking (InfiniBand or RoCE) and storage; Availability date and minimum term; Your indicative price per GPU-hour) - What we do not do: So there are no surprises on either side. (We do not publish your capacity, prices or name; We do not charge to list or hold your capacity; We do not offer capacity we have not verified; We never share your details with a buyer without your go-ahead and an NDA) ### Tell us what you have The more precise you are, the faster we can tell you whether we have a buyer for it. Everything you share is confidential. ### What partners ask **Who can partner with FlexiCloud?** Owners and operators of dedicated GPU capacity: GPU cloud providers, data centres, and organisations with spare nodes. If the hardware is leased or brokered, tell us up front. We can still talk, but we verify who really controls it before offering it to anyone. **What does it cost to partner?** Nothing to list or hold your capacity. FlexiCloud earns a commission only when a deal closes, so we are paid when you are. The commission terms are agreed with you directly. **Will you publish my capacity or prices?** No. We never publish partner names, inventory or prices. A buyer sees your details only under NDA and with your go-ahead. **What is the minimum I can offer?** One full 8-GPU node, as a rule. Buyers contract for whole nodes because the eight GPUs share NVLink. **Which GPUs and regions do you want?** NVIDIA H100, H200, B200 and B300 today, and Vera Rubin when it is available. India is the first priority. Capacity in the EU and the US is welcome too. **Does FlexiCloud buy or hold capacity?** No. We do not own hardware. We bring buyers to providers, and we are paid a commission when a deal closes. **What happens after I submit?** The founder reads every submission personally. If there is a fit with the buyers we are talking to, we reply with an NDA and set up a call. If there is not, we tell you that too. ### More on the model Why buyers come to us instead of going direct, and how the GPU generations compare. ### Ready to put your capacity to work? It takes about three minutes to tell us what you have. ## Pricing URL: https://flexicloud.co/pricing Dedicated 8-GPU NVIDIA nodes from $2.95 per GPU-hour on our longest term. Website hosting too: managed cloud from $10/month, shared hosting from ₹1,750/year. Last modified: 2026-10-18 ### NVIDIA GPU pricing, from $2.95 per GPU-hour Dedicated 8-GPU B300, B200, H200 and H100 nodes for India, the EU or the US, sourced across more than ten partner providers. The rates shown are floor rates on our longest term. Looking for website hosting? Managed cloud and shared hosting plans are lower on this page. ### Which prices do you need? This page has two price lists: GPU servers first, then website hosting. Each has its own terms, guarantees and questions. - Dedicated 8-GPU NVIDIA nodes: For AI training and inference. Per GPU-hour, in USD, before tax. Floor rates on our longest term; the one-year minimum is priced higher. Billed upfront. (H100 from $2.95 per GPU-hour; H200 from $3.35 per GPU-hour; B200 from $3.75 per GPU-hour; B300 from $4.50 per GPU-hour) - Managed cloud and shared hosting: For websites and apps, on managed CPU servers or shared plans, billed monthly or yearly. (Managed cloud from $10 a month, in USD, before tax; Shared hosting from ₹1,750 a year, in INR, excluding 18% GST; Free migration; 30-day money back on shared hosting and monthly managed cloud plans, for first-time customers) ### GPU compute, priced by the GPU-hour Dedicated NVIDIA SXM bare-metal nodes for training and inference, quoted for India, the EU or the US. Every node has 8 GPUs, so a node costs 8 times the GPU-hour rate. The rates shown are floor rates on our longest term; one-year terms, the minimum, are priced higher. Provider: NVIDIA GPU nodes - 8-GPU HGX B300 node: from USD 4.50 per GPU-hour. Blackwell Ultra, the most memory per GPU we offer, for the largest models and longest context. Dedicated bare metal, single tenant; we keep the hardware running for your term. (Floor on our longest term: $36.00 per node-hour, about $26,300 per 730-hour month; 8 × NVIDIA B300 SXM, 288 GB HBM3e each; NVLink across the node, 2.3 TB HBM total; Minimum one node on a one-year term, priced above this floor; Quoted for India, the EU or the US; Hardware monitored and kept running for you) - 8-GPU HGX B200 node: from USD 3.75 per GPU-hour. Blackwell, the cost-effective node for large-model training and inference that fits in 180 GB per GPU. Dedicated bare metal, single tenant; we keep the hardware running for your term. (Floor on our longest term: $30.00 per node-hour, about $21,900 per 730-hour month; 8 × NVIDIA B200 SXM, 180 GB HBM3e each; NVLink across the node, 1.44 TB HBM total; Minimum one node on a one-year term, priced above this floor; Quoted for India, the EU or the US; Hardware monitored and kept running for you) - 8-GPU HGX H200 node: from USD 3.35 per GPU-hour. Hopper with 141 GB per GPU, for long context and models that outgrow 80 GB. Dedicated bare metal, single tenant; we keep the hardware running for your term. (Floor on our longest term: $26.80 per node-hour, about $19,600 per 730-hour month; 8 × NVIDIA H200 SXM, 141 GB HBM3e each; NVLink across the node, 1.1 TB HBM total; Minimum one node on a one-year term, priced above this floor; Quoted for India, the EU or the US; Hardware monitored and kept running for you) - 8-GPU HGX H100 node: from USD 2.95 per GPU-hour. The proven Hopper workhorse with 80 GB per GPU, at our lowest rate. Dedicated bare metal, single tenant; we keep the hardware running for your term. (Floor on our longest term: $23.60 per node-hour, about $17,200 per 730-hour month; 8 × NVIDIA H100 SXM, 80 GB HBM3 each; NVLink across the node, 640 GB HBM total; Minimum one node on a one-year term, priced above this floor; Quoted for India, the EU or the US; Hardware monitored and kept running for you) ### Get a GPU quote Leave a name and a number or email. We call you back with pricing for your GPU, region and term, usually within one working day. ### GPU node pricing: common questions **What does a whole 8-GPU node cost?** Eight times the per GPU-hour rate. At our floor rates, on our longest term, a node costs from $23.60 an hour for H100, $26.80 for H200, $30.00 for B200 and $36.00 for B300: about $17,200, $19,600, $21,900 and $26,300 over a 730-hour month. One-year terms are priced higher. All rates are in USD, before tax. Your exact rate depends on region, term and number of nodes, and is confirmed in writing before you commit. **What does the "from" price mean for GPU nodes?** It is our floor rate, on our longest term. A one-year term, the minimum, is priced higher than the floor. Your rate depends on region, term length and number of nodes, and is confirmed in writing before you commit. **What is the minimum GPU order?** One full 8-GPU node on a one-year minimum term, as dedicated single-tenant bare metal. **How are GPU nodes billed?** Upfront for the term. A GPU node is billed in advance for its committed term, and the finer details are agreed case by case and confirmed in writing before you commit. Rates are in USD and before tax; any applicable taxes are added to the invoice as a separate charge. **What does the GPU-hour rate include?** The dedicated 8-GPU node as bare metal with network access, and our work keeping the hardware running for your term: monitoring, failure handling and support. You run the software on it, including the operating system, drivers, libraries and orchestration. The managed hosting services in the hosting part of this page are not part of a GPU node. **Does the 30-day money-back guarantee cover GPU nodes?** No. A GPU node is sold on a committed term, which is billed in full and is not covered by the 30-day money-back guarantee, as set out in our Terms. ### Pricing a GPU node: worth reading ### Website hosting: managed cloud and shared plans Hosting for websites and apps, on managed CPU servers or shared plans, billed monthly or yearly. Everything from here down is for hosting only; GPU prices and terms are at the top of this page. ### Managed cloud hosting: pick a provider, then a size CPU servers for websites and apps, such as WordPress and WooCommerce, separate from the GPU nodes above. Same management on every provider, with no management fee on top: provisioning, tuning, patching and 24×7 monitoring are included. Pick the provider whose regions suit your audience; we will recommend one if you are unsure. Provider: DigitalOcean - Micro: USD12/month, USD144/year (1 vCPU; 1 GB RAM; 25 GB SSD; 1TB bandwidth; Unlimited websites) - Small: USD24/month, USD288/year (1 vCPU; 2 GB RAM; 55 GB SSD; 2TB bandwidth; Unlimited websites) - Medium: USD48/month, USD576/year (2 vCPU; 4 GB RAM; 80 GB SSD; 3TB bandwidth; Unlimited websites) - Large: USD96/month, USD1,152/year (4 vCPU; 8 GB RAM; 160 GB SSD; 4TB bandwidth; Unlimited websites) Provider: Vultr - Micro: USD10/month, USD100/year (1 vCPU; 1 GB RAM; 25 GB SSD; 1TB bandwidth; Unlimited websites) - Small: USD20/month, USD200/year (1 vCPU; 2 GB RAM; 55 GB SSD; 2TB bandwidth; Unlimited websites) - Medium: USD40/month, USD400/year (2 vCPU; 4 GB RAM; 80 GB SSD; 3TB bandwidth; Unlimited websites) - Large: USD80/month, USD800/year (4 vCPU; 8 GB RAM; 160 GB SSD; 4TB bandwidth; Unlimited websites) Provider: Hetzner - Micro: USD15/month, USD180/year (2 vCPU; 4 GB RAM; 40 GB SSD; 20TB bandwidth; Unlimited websites) - Medium: USD30/month, USD360/year (4 vCPU; 8 GB RAM; 160 GB SSD; 20TB bandwidth; Unlimited websites) - Large: USD60/month, USD720/year (8 vCPU; 16 GB RAM; 240 GB SSD; 20TB bandwidth; Unlimited websites) Provider: Linode - Micro: USD12/month, USD144/year (1 vCPU; 1 GB RAM; 25 GB SSD; 1TB bandwidth; Unlimited websites) - Small: USD24/month, USD288/year (1 vCPU; 2 GB RAM; 55 GB SSD; 2TB bandwidth; Unlimited websites) - Medium: USD48/month, USD576/year (2 vCPU; 4 GB RAM; 80 GB SSD; 3TB bandwidth; Unlimited websites) - Large: USD96/month, USD1,152/year (4 vCPU; 8 GB RAM; 160 GB SSD; 4TB bandwidth; Unlimited websites) ### What we look after on managed cloud servers This covers managed cloud servers. Shared hosting includes what is listed on each plan, and GPU nodes are bare metal you run yourself; see the GPU questions at the top of this page. - Provisioned and tuned: We size the server, build the stack, and tune LiteSpeed, PHP, Redis and MySQL for your workload. You get a running site, not an empty Ubuntu box. - Patched and hardened: OS and stack updates, firewall rules, SSL renewal and malware scanning are ours to keep current, including the ones that land at 2am. - Watched around the clock: Uptime, CPU, memory and disk are monitored continuously. When something moves the wrong way, an engineer is already looking at it. - Backups you can restore: Offsite backups with one-click restore, with storage billed at $0.033 per GB per month. We test restores, because a backup you have never restored is not a backup. - Scaled without downtime: Traffic spike or a sale coming up? We resize the server or add capacity, usually while the site stays up. - Supported by people: Real engineers on chat, ticket and phone, in your timezone, not eleven hours behind it. ### Shared hosting for smaller sites Shared hosting with the Webuzo control panel, for sites that do not need a server of their own yet. Priced in Indian rupees. Move up to managed cloud any time, free. Provider: Shared hosting - Starter 2: INR1,750/year. One small site, annual term. (1 GB disk space; 1 database; 1 domain; Unlimited bandwidth; Free SSL) - Business 1: INR500/month, INR5,000/year. A single growing site. (5 GB disk space; 5 websites; 5 databases; Unlimited bandwidth; Free SSL) - Business 2: INR900/month, INR9,000/year. Several sites under one account. (10 GB disk space; Unlimited databases; 9 domains; Unlimited bandwidth; Free SSL) - Expert 1: INR1,250/month, INR15,000/year. Busier sites needing more room. (15 GB disk space; Unlimited websites; Unlimited databases; Unlimited bandwidth; Free SSL) - Expert 2: INR1,875/month, INR22,500/year. The most a shared server should carry. (25 GB disk space; Unlimited databases; Unlimited domains; Unlimited bandwidth; Free SSL) ### Managed cloud and shared hosting: billing questions **Is there a management fee on managed cloud hosting?** No. The managed cloud price covers the server and our management of it. There is no per-site or per-seat charge on top. **What does the 30-day money-back guarantee cover?** Shared hosting and monthly managed cloud plans, for first-time customers, as set out in our Terms. If the service is not right for you within the first 30 days, we refund it. GPU nodes are sold on a committed term and are not covered. **Do hosting prices include tax?** No. Managed cloud prices are in USD and before tax. Shared hosting prices are in INR and exclude 18% GST. Any applicable taxes are added to your invoice as a separate charge. **How do I pay for hosting?** Credit and debit cards, UPI, netbanking and NEFT. International customers can pay by card, ACH or wire transfer. **What happens if I outgrow my hosting plan?** On managed cloud we resize the server, usually without downtime, and you move onto the next plan’s price from that billing cycle. On shared hosting we move you up a plan, or onto managed cloud, free. **Are backups extra on managed cloud?** Yes. Offsite backup storage is billed separately, at $0.033 per GB per month. ### What a hosting plan actually buys ### Not sure what to order? For GPUs, tell us the model, number of nodes, region and term. For hosting, send us your traffic numbers. We size it and reply in writing. ## Privacy Policy URL: https://flexicloud.co/privacy-policy How FlexiCloud collects, uses and protects your personal data when you use our website and hosting services. Last modified: 2026-10-14 ### Privacy Policy How we collect, use and protect your personal data when you use FlexiCloud. FlexiCloud Hosting (“us”, “we”, or “our”) operates the flexicloud.co website (the “Service”). This page informs you of our policies regarding the collection, use, and disclosure of personal data when you use our Service and the choices you have associated with that data. We use your data to provide and improve the Service. By using the Service, you agree to the collection and use of information in accordance with this policy. Unless otherwise defined in this Privacy Policy, terms used in this Privacy Policy have the same meanings as in our Terms and Conditions. Information Collection And Use We collect several different types of information for various purposes to provide and improve our Service to you. Personal Data While using our Service, we may ask you to provide us with certain personally identifiable information that can be used to contact or identify you (“Personal Data”). Personally identifiable information may include, but is not limited to: Email address First name and last name Phone number Address, State, Province, ZIP/Postal code, City Cookies and Usage Data Usage Data We may also collect information about how the Service is accessed and used (“Usage Data”). This Usage Data may include information such as your computer’s Internet Protocol address (e.g. IP address), browser type, browser version, the pages of our Service that you visit, the time and date of your visit, the time spent on those pages, unique device identifiers and other diagnostic data. Tracking & Cookies Data We use cookies and similar tracking technologies to track the activity on our Service and hold certain information. Cookies are files with small amount of data which may include an anonymous unique identifier. Cookies are sent to your browser from a website and stored on your device. Tracking technologies also used are beacons, tags, and scripts to collect and track information and to improve and analyze our Service. You can instruct your browser to refuse all cookies or to indicate when a cookie is being sent. However, if you do not accept cookies, you may not be able to use some portions of our Service. Examples of Cookies we use: Session Cookies: We use Session Cookies to operate our Service. Preference Cookies: We use Preference Cookies to remember your preferences and various settings. Security Cookies: We use Security Cookies for security purposes. Use Of Data FlexiCloud Hosting uses the collected data for various purposes: To provide and maintain the Service To notify you about changes to our Service To allow you to participate in interactive features of our Service when you choose to do so To provide customer care and support To provide analysis or valuable information so that we can improve the Service To monitor the usage of the Service To detect, prevent and address technical issues Transfer Of Data Your information, including Personal Data, may be transferred to, and maintained on, computers located outside of your state, province, country or other governmental jurisdiction where the data protection laws may differ than those from your jurisdiction. If you are located outside India and choose to provide information to us, please note that we transfer the data, including Personal Data, to India and process it there. Your consent to this Privacy Policy followed by your submission of such information represents your agreement to that transfer. FlexiCloud Hosting will take all steps reasonably necessary to ensure that your data is treated securely and in accordance with this Privacy Policy and no transfer of your Personal Data will take place to an organization or a country unless there are adequate controls in place including the security of your data and other personal information. Disclosure Of Data Legal Requirements FlexiCloud Hosting may disclose your Personal Data in the good faith belief that such action is necessary to: To comply with a legal obligation To protect and defend the rights or property of FlexiCloud Hosting To prevent or investigate possible wrongdoing in connection with the Service To protect the personal safety of users of the Service or the public To protect against legal liability Security Of Data The security of your data is important to us, but remember that no method of transmission over the Internet, or method of electronic storage is 100% secure. While we strive to use commercially acceptable means to protect your Personal Data, we cannot guarantee its absolute security. Service Providers We may employ third party companies and individuals to facilitate our Service (“Service Providers”), to provide the Service on our behalf, to perform Service-related services or to assist us in analyzing how our Service is used. These third parties have access to your Personal Data only to perform these tasks on our behalf and are obligated not to disclose or use it for any other purpose. Analytics We may use third-party Service Providers to monitor and analyze the use of our Service. Google Analytics: Google Analytics is a web analytics service offered by Google that tracks and reports website traffic. Google uses the data collected to track and monitor the use of our Service. This data is shared with other Google services. Google may use the collected data to contextualize and personalize the ads of its own advertising network. You can opt-out of having made your activity on the Service available to Google Analytics by installing the Google Analytics opt-out browser add-on. The add-on prevents the Google Analytics JavaScript (ga.js, analytics.js, and dc.js) from sharing information with Google Analytics about visits activity. For more information on the privacy practices of Google, please visit the Google Privacy & Terms web page. Links To Other Sites Our Service may contain links to other sites that are not operated by us. If you click on a third party link, you will be directed to that third party’s site. We strongly advise you to review the Privacy Policy of every site you visit. We have no control over and assume no responsibility for the content, privacy policies or practices of any third party sites or services. Children’s Privacy Our Service does not address anyone under the age of 18 (“Children”). We do not knowingly collect personally identifiable information from anyone under the age of 18. If you are a parent or guardian and you are aware that your Children has provided us with Personal Data, please contact us. If we become aware that we have collected Personal Data from children without verification of parental consent, we take steps to remove that information from our servers. Governing Law and Jurisdiction This Privacy Policy and any dispute arising out of or relating to it are governed by the laws of India. You irrevocably consent to the exclusive jurisdiction and venue of the courts in Kochi, Kerala, India. Changes To This Privacy Policy We may update our Privacy Policy from time to time. We will notify you of any changes by posting the new Privacy Policy on this page. We will let you know via email and/or a prominent notice on our Service, prior to the change becoming effective and update the “effective date” at the top of this Privacy Policy. You are advised to review this Privacy Policy periodically for any changes. Changes to this Privacy Policy are effective when they are posted on this page. Contact Us If you have any questions about this Privacy Policy, please contact us by email: support@flexicloud.in ## FlexiCloud Panel URL: https://flexicloud.co/products/flexicloud-panel Deploy sites, restore backups, add users and read invoices across every server you run with FlexiCloud, in one panel, without SSH. Last modified: 2026-10-18 ### One panel for every server you run Whichever cloud your servers are on, they appear in the same place, with the same deploys, the same backups and one invoice. ### The things you would otherwise open a ticket for - Deploy a site: Add a site to any server, with the stack configured and SSL issued automatically. - Backups and restores: Restore any backup from the same screen, without raising a ticket or waiting for us. - Atomic deployments: Deploys either complete or they do not happen. Visitors never see a half-updated site. - Users and access: Give a client or a developer access to one site without handing over the server. - Resource visibility: CPU, memory and disk per server, so you can see a problem coming. - Support in context: Raise a ticket from the site it is about, with the details already attached. ### Every site, every cloud, one list Servers on DigitalOcean, Vultr, Hetzner and Linode sit side by side. The provider becomes an implementation detail. ### The panel in practice ### Want to see it? A demo takes fifteen minutes and we will use your real setup as the example. ## Agency Hosting URL: https://flexicloud.co/solutions/agency-hosting Run every client on its own server or pack them onto shared infrastructure, your call per client. Give clients logins without your billing. Last modified: 2026-10-18 ### Many client sites, one panel, one invoice Run every client on its own server or pack them onto shared infrastructure. Your call, per client. Give clients their own logins without giving them your billing. ### What you get - Every client, one dashboard: All servers and all sites in a single panel, grouped by client, with per-site access control. - One bill, resold at your rate: Consolidated invoicing from us. What you charge your clients is between you and them. - Migrations we do for you: Bringing a portfolio of client sites across is our work, not a weekend of yours. Free, and scheduled around your clients. - Support that answers as you: White-label support available, so an escalation to us does not look like an escalation to your client. ### Everything in one panel Sites, servers, backups, users and invoices, without a support ticket and without touching SSH. ### Getting you here Free, scheduled around you, with a fallback the whole way. - Tell us what you run: Send us the current host, the stack and a rough traffic figure. No access needed yet. - We size and build: We recommend a provider and plan, then build the server and copy the site across. - You test on a temp URL: The site runs on a temporary address until you have clicked through it and signed off. - We switch DNS: Cutover happens at a time you choose, with the old host kept warm as a fallback. ### Useful if you manage client sites ### Agency Hosting that someone else operates. Tell us what you are running today. ## E-commerce Hosting URL: https://flexicloud.co/solutions/ecommerce-hosting A shop is not a brochure site. Carts and checkouts cannot be cached and traffic spikes. A hosting stack tuned for e-commerce that stays up. Last modified: 2026-10-18 ### Hosting for stores that cannot afford a bad checkout A shop is not a brochure site. Carts and checkouts cannot be cached, traffic arrives in spikes, and every minute of downtime is money. This is a stack tuned for that. ### What you get - Tuned for carts, not just pages: Product pages cache beautifully; carts and checkouts cannot. We tune object caching, PHP workers and the database for the dynamic half of a store, which is the half that takes payments. - Ready before the spike: Tell us a sale or campaign is coming and we resize ahead of it, then scale back afterwards. Traffic you planned for should never be the reason a store falls over. - Someone watching at checkout: Monitoring is on the paths that matter, and an engineer is on call. You should not learn that checkout broke from a customer email the next morning. - Secure by default: Free SSL kept renewed, firewall rules, malware scanning and isolated environments. Card data belongs with your payment gateway, and everything around it is hardened. ### Everything in one panel Sites, servers, backups, users and invoices, without a support ticket and without touching SSH. ### Getting you here Free, scheduled around you, with a fallback the whole way. - Tell us what you run: Send us the current host, the stack and a rough traffic figure. No access needed yet. - We size and build: We recommend a provider and plan, then build the server and copy the site across. - You test on a temp URL: The site runs on a temporary address until you have clicked through it and signed off. - We switch DNS: Cutover happens at a time you choose, with the old host kept warm as a fallback. ### More on running a store ### E-commerce Hosting that someone else operates. Tell us what you are running today. ## Enterprise Hosting URL: https://flexicloud.co/solutions/enterprise-hosting Isolated environments, defined SLAs, and a team that knows your architecture, for the sites where an hour of downtime is a board-level conversation. Last modified: 2026-10-18 ### Dedicated infrastructure, named engineers Isolated environments, defined SLAs, and a team that knows your architecture, for the sites where an hour of downtime is a board-level conversation. ### What you get - Isolated environments: Dedicated servers, private networking and isolated environments, kept apart from any other customer. - Defined SLAs: Response and resolution commitments in writing, with escalation paths that reach a human. - Architecture review: We look at how the application is actually built before recommending what it should run on. - Named engineers: The same people every time, who already know your setup when something breaks. ### Everything in one panel Sites, servers, backups, users and invoices, without a support ticket and without touching SSH. ### Getting you here Free, scheduled around you, with a fallback the whole way. - Tell us what you run: Send us the current host, the stack and a rough traffic figure. No access needed yet. - We size and build: We recommend a provider and plan, then build the server and copy the site across. - You test on a temp URL: The site runs on a temporary address until you have clicked through it and signed off. - We switch DNS: Cutover happens at a time you choose, with the old host kept warm as a fallback. ### Planning for scale ### Enterprise Hosting that someone else operates. Tell us what you are running today. ## Managed WordPress Hosting URL: https://flexicloud.co/solutions/managed-wordpress-hosting Object caching, page caching, PHP workers and database tuning set up correctly for WordPress, plus free migration and an engineer who knows what a 502 means. Last modified: 2026-10-18 ### WordPress, tuned by people who run it every day Object caching, page caching, PHP workers and database tuning set up correctly for WordPress, plus free migration and an engineer who knows what a 502 means. ### What you get - A stack built for WordPress: LiteSpeed with LSAPI PHP, Redis object cache and OPcache, tuned to your plugin load rather than left at defaults. LiteSpeed Cache serves pages from the server itself instead of rebuilding them in PHP. - Your .htaccess just works: LiteSpeed reads .htaccess natively, so redirects, rewrites and plugin rules come across from your old host unchanged. Nothing to translate, nothing to re-test. - Security that assumes WordPress: Login hardening, file permission enforcement, malware scanning and automatic core and plugin updates where you want them. - Support that reads the error log: Tell us the site is slow and we will find out why, rather than sending you a link to a documentation page. ### Everything in one panel Sites, servers, backups, users and invoices, without a support ticket and without touching SSH. ### Getting you here Free, scheduled around you, with a fallback the whole way. - Tell us what you run: Send us the current host, the stack and a rough traffic figure. No access needed yet. - We size and build: We recommend a provider and plan, then build the server and copy the site across. - You test on a temp URL: The site runs on a temporary address until you have clicked through it and signed off. - We switch DNS: Cutover happens at a time you choose, with the old host kept warm as a fallback. ### More on running WordPress ### Managed WordPress Hosting that someone else operates. Tell us what you are running today. ## Startup Hosting URL: https://flexicloud.co/solutions/startup-hosting Managed cloud hosting for startups on DigitalOcean, Vultr, Hetzner or Linode. Start from $10 a month, scale the same server as you grow, free migration. Last modified: 2026-10-26 ### Start on one server. Do not replatform at ten. Begin on the smallest cloud instance that works and scale the same infrastructure as you grow: no migration, no rewrite, no surprise enterprise pricing tier. ### What you get - Start small, honestly: A $10–12 instance is a genuinely fine place to launch. We will tell you when it stops being one. - Scale without a migration: Resize the server as traffic grows. The stack, the URLs and the deploy process stay the same. - Not just WordPress: Laravel, Django, Node and static builds run here too, with the same management and monitoring. - No infrastructure hire needed: Your engineers ship product. Server patching, monitoring and incident response are ours. ### Everything in one panel Sites, servers, backups, users and invoices, without a support ticket and without touching SSH. ### Getting you here Free, scheduled around you, with a fallback the whole way. - Tell us what you run: Send us the current host, the stack and a rough traffic figure. No access needed yet. - We size and build: We recommend a provider and plan, then build the server and copy the site across. - You test on a temp URL: The site runs on a temporary address until you have clicked through it and signed off. - We switch DNS: Cutover happens at a time you choose, with the old host kept warm as a fallback. ### Sizing and spending sensibly ### Startup Hosting that someone else operates. Tell us what you are running today. ## Terms & Conditions URL: https://flexicloud.co/terms-conditions The terms governing your use of FlexiCloud managed cloud hosting, shared hosting and GPU compute: payments, refunds, acceptable use, availability and liability. Last modified: 2026-09-30 ### Terms & Conditions The terms that govern your use of FlexiCloud. These Terms of Service ("Terms", "ToS") govern your access to and use of FlexiCloud's websites, hosting and compute services (the "Services"). By signing up for or using the Services, you agree to these Terms. If you do not agree, do not use the Services. These Terms work alongside our Privacy Policy, which explains how we handle your personal data. Terms used here have the same meaning as in that policy unless defined otherwise. Payments and refund All payments are made in advance. You will receive an email when a payment is due, and can pay through our billing system by card, UPI, netbanking, NEFT or bank transfer. Listed fees do not include applicable taxes. Any applicable GST, cess or other taxes will be added to your invoice as a separate charge. All fees are non-refundable once paid unless otherwise stated in these Terms. If a payment is not made by its due date, we may levy a late fee of 10% of the invoice amount and suspend the Service. A suspended account can be reinstated only after all outstanding dues and late fees are paid, within a reasonable time at our discretion. Non-payment for two consecutive months will result in cancellation of the Service and deletion of the associated data. Refund We offer a 30-day money-back guarantee to new customers on shared hosting and monthly managed cloud plans, if you are not satisfied with the Service. This applies to first-time accounts only. If you have or have had an account with us, a further account is not treated as a new-customer account and is not eligible for new-customer offers or refunds. If your account is terminated for a violation of these Terms, you are not eligible for a refund. Refund eligibility and exclusions Only first-time accounts are eligible for the money-back guarantee. GPU compute and any service sold on a fixed term or minimum commitment (see "GPU compute" below) are not covered by the 30-day guarantee. Committed terms are billed for the full term. Dedicated resources, administrative fees, and installation fees for custom software are non-refundable. Domain name purchases are non-refundable. If a domain was ordered with a hosting package, any refund is issued after deducting the domain charge at market or purchase value, whichever is higher. Refunds on promotional offers, discounted sales and bundles are at our sole discretion. The Services In exchange for the fees you pay in advance, FlexiCloud provides the Services described for the plan you select at signup, subject to these Terms and any other agreement relating to the Services. We offer: Managed cloud hosting on third-party cloud providers (DigitalOcean, Vultr, Hetzner and Linode), where we provision, tune, secure, patch and monitor the server on the cloud you choose. Shared hosting administered through the Webuzo control panel, for smaller websites. GPU compute - dedicated bare-metal GPU nodes for AI training and inference. Where the Services run on a third-party cloud provider's infrastructure, that provider's own terms may also apply to the underlying resources. GPU compute GPU compute is sold as dedicated, single-tenant bare-metal nodes on a minimum term (currently one year) and priced per GPU-hour. For these Services: The minimum order is one node, and the minimum commitment is the stated term. The committed term is billed in full and is not covered by the 30-day money-back guarantee. FlexiCloud delivers the node as bare metal with network access and keeps the hardware running - monitoring, failure handling and support - for the term. You are responsible for the software you run on it, including the operating system, drivers, libraries and orchestration. Availability is confirmed per region before an order is accepted. Final pricing depends on region, term length and the number of nodes, and is confirmed in writing before you commit. Backup policy We take automated backups of managed cloud and shared hosting accounts and retain them for a rolling recent period, which you can restore through your control panel or by contacting support. Backups are provided as a convenience and not as a guarantee. Like any storage system, backups can fail, and some large files may be excluded. You remain responsible for maintaining your own independent backups of your content. FlexiCloud is not liable for loss or corruption of your content, or for any interruption caused by it. We strongly recommend you keep an additional copy of your website and data outside our systems. Acceptable usage policy FlexiCloud maintains a high standard of service that many customers depend on. We will not provide Services to anyone using them for: Hacking, including penetrating or attempting to access, without authorization, another computer or network. Port scans, stealth scans and credit-card "phishing" are prohibited. Hosting files or data that infringe another's copyright or intellectual property rights. Spamming, or sending bulk unsolicited email, including maintaining an open SMTP relay. We may refuse or terminate service on reasonable indication of spamming of any kind. Content that violates another's privacy or publicity rights; hate speech; or content that is obscene, abusive, libelous or defamatory. Hosting, storing or distributing child sexual abuse material. Any such content is reported to the appropriate law-enforcement agencies and the account is suspended immediately without notice. Bulk file storage or backup not tied to an active hosting account. All Services may be used for lawful purposes only. You agree to indemnify and hold FlexiCloud harmless from any claim resulting from your use of the Services. Prohibited material includes, without limitation: IRC bots, proxy scripts and anonymizers, pirated software or "warez", file-dump or mirror services, HYIP and investment-scheme sites, unlicensed sale of controlled substances, AutoSurf sites, lottery and gambling sites, hacking-focused archives or tools, IP scanners, brute-force programs, mail bombers and spam scripts. Forums or sites that distribute or link to warez are prohibited. FlexiCloud systems and networks are provided for authorized customer use and may be monitored for lawful purposes, including security and management of the system. Use of the Services constitutes consent to such monitoring. We reserve the right to refuse service to anyone, and to remove or disable any material that, in our judgment, is illegal or violates these Terms, with or without notice. You are responsible for the security of scripts and software installed under your account, for setting appropriate file permissions, and for using strong credentials. You are responsible for all activity under your account, including any compromise of your credentials. Failure to respond to our abuse team within 48 hours may result in suspension or termination. If in doubt about the acceptability of your site or service, contact us at abuse@flexicloud.in. Spam policy We take a zero-tolerance stance on unsolicited email, bulk emailing and spam. "Safe lists" and "double opt-in" lists are treated as spam. Any account that sends spam may be terminated with or without notice. Sites advertised via spam ("spamvertised") may not be hosted with us, including spam sent by fax, email, instant messaging or newsgroups. Any account that causes our IP space to be blacklisted may be suspended or terminated immediately. We reserve the right to require changes to, or disable, any site, account or component that does not comply with our policies, and to charge a clean-up fee at our discretion for accounts used to send unsolicited email. Intellectual property and DMCA We respect the intellectual property of others and ask you to do the same. In accordance with the Digital Millennium Copyright Act and applicable Indian law, including the Information Technology Act, 2000, we may suspend or terminate the accounts of users found to infringe copyright, and of repeat infringers. If you believe your work has been copied in a way that constitutes infringement, please provide our copyright agent at abuse@flexicloud.in with: (1) a signature of the person authorized to act for the rights owner; (2) a description of the copyrighted work claimed to be infringed; (3) the location of the infringing material on the site; (4) your address, telephone number and email; (5) a statement that you have a good-faith belief the use is not authorized; and (6) a statement, under penalty of perjury, that the information is accurate and that you are, or are authorized to act for, the rights owner. Service availability We will endeavour to provide the Services 24 hours a day, 7 days a week for as long as you have paid for them. However, the Services may occasionally be unavailable due to equipment malfunction, maintenance or repairs, or causes beyond our control such as power failures, telecommunication failures, network attacks or traffic. Due to the nature of the technology, occasional unavailability cannot be entirely avoided, and we do not promise uninterrupted service. Indemnification If FlexiCloud is sued or threatened with a lawsuit by a third party because of your use of the Services, you agree to indemnify FlexiCloud: to reimburse any amount FlexiCloud pays to satisfy a judgment or settle the matter, together with FlexiCloud's reasonable legal fees and other costs. Disclaimer of warranties FLEXICLOUD EXPRESSLY DISCLAIMS ALL WARRANTIES OF ANY KIND, WHETHER EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE IMPLIED WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. THE SERVICES ARE PROVIDED ON AN "AS IS" AND "AS AVAILABLE" BASIS. FLEXICLOUD MAKES NO WARRANTY THAT THE SERVICES WILL MEET YOUR REQUIREMENTS, OR THAT THEY WILL BE UNINTERRUPTED, TIMELY, SECURE OR ERROR-FREE, OR THAT DEFECTS WILL BE CORRECTED. Some jurisdictions do not allow the disclaimer of implied warranties, in which case the foregoing may not apply to you. Limitation of liability IN NO EVENT SHALL FLEXICLOUD BE LIABLE TO YOU OR ANY OTHER PERSON FOR ANY INDIRECT, INCIDENTAL, CONSEQUENTIAL, SPECIAL, EXEMPLARY OR PUNITIVE DAMAGES, INCLUDING LOSS OF PROFIT OR GOODWILL, WHETHER BASED ON CONTRACT, TORT (INCLUDING NEGLIGENCE), BREACH OF WARRANTY, ANY BREACH OF THESE TERMS, YOUR INABILITY TO USE THE SERVICES, OR YOUR LOSS OF DATA, EVEN IF FLEXICLOUD HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH DAMAGES. You agree that our maximum aggregate liability shall not exceed the total amount you paid for the Services in dispute. Some jurisdictions do not allow such limitations, in which case our liability is limited to the fullest extent permitted by law. Third-party services The Services may run on, or integrate with, third-party products and services, including cloud infrastructure providers and GPU hardware partners. FlexiCloud assumes no responsibility or liability for third-party products and services, which may be governed by their own terms. Data protection Our handling of your personal data is described in our Privacy Policy. We process personal data in accordance with applicable Indian law, including the Digital Personal Data Protection Act, 2023. Choice of law and disputes These Terms are governed by the laws of India. You irrevocably consent to the exclusive jurisdiction and venue of the courts in Kochi, Kerala, India for all disputes arising out of or relating to these Terms or the Services. Entire agreement These Terms are the entire agreement regarding your use of the Services and supersede any prior or concurrent communications or agreements between you and FlexiCloud regarding the Services. Changes to these Terms As our business and the industry evolve, these Terms may change. We will post changes on this page and update the effective date. Your continued use of the Services after a change means you accept the updated Terms. Contact Questions about these Terms can be sent to us via the contact page (https://flexicloud.co/contact). Abuse and copyright notices should go to abuse@flexicloud.in. ## Articles - [Vera Rubin explained: what changes and when it matters to buyers](https://flexicloud.co/blogs/nvidia-vera-rubin-explained) (2026-10-24): NVIDIA Vera Rubin is shipping. Here is what changes from Blackwell, what analysts estimate the hardware costs, how the cable-free trays are built, whether it is the "wireless" server Jensen Huang talked about, how much faster it really is, and when it matters for your plans. - [Hosting your GPUs in India: latency, power, support and skills](https://flexicloud.co/blogs/hosting-gpus-in-india) (2026-10-23): More European and US companies are looking at India for their GPUs. Here are honest answers on latency, internet, power, support, skills and whether India can meet the standards of cutting-edge AI companies. - [GB200 vs B200, GB300 vs B300: what Grace Blackwell and NVL72 mean](https://flexicloud.co/blogs/gb200-vs-b200-gb300-vs-b300-nvl72-explained) (2026-10-16): B200 and B300 are GPUs; GB200 and GB300 pair them with NVIDIA’s Grace CPU and usually come as 72-GPU NVL72 racks. What that means, side by side, and who needs which. - [PCIe, SXM, HGX, DGX and NVL72: NVIDIA GPU form factors explained](https://flexicloud.co/blogs/nvidia-gpu-form-factors-pcie-sxm-hgx-dgx-nvl) (2026-10-16): PCIe, SXM, HGX, DGX, NVL and NVL72 explained in plain language, with a table of which form the H100, H200, B200, B300, GB200, GB300, RTX PRO 6000 and Vera Rubin come in. - [Bare metal vs cloud GPUs: who can see your data?](https://flexicloud.co/blogs/bare-metal-vs-cloud-gpu-security) (2026-10-15): On a cloud GPU, the cloud layer can in principle read and copy your machine. Bare metal removes that layer, but you still trust the operator. Here is who can see your data, and what to ask before you rent. - [RTX PRO 6000 vs H100: which GPU is right for your business?](https://flexicloud.co/blogs/rtx-pro-6000-vs-h100) (2026-10-13): The H100 is the faster GPU for training AI and for giving every user fast replies. The RTX PRO 6000 has more memory on one card and suits video, graphics and teams sharing a GPU. Here is how to choose. - [Renting GPUs: why going through FlexiCloud beats going direct](https://flexicloud.co/blogs/why-rent-gpus-through-flexicloud) (2026-10-06): We partner with more than ten GPU providers in India, the EU and the US, so you do not have to. Here is what that gets you: better pricing, priority on scarce GPUs, live inventory across datacentres and regions, privacy, and one honest point of contact instead of ten sales teams. - [Hopper vs Blackwell: NVIDIA’s GPU generations compared](https://flexicloud.co/blogs/nvidia-hopper-vs-blackwell) (2026-10-02): NVIDIA’s Hopper (H100, H200) and Blackwell (B200, B300) generations compared: memory, architecture, interconnect, power and price, and a practical guide to which one actually fits your AI workload. - [Renting a multi-node B300 cluster: what to check before you commit](https://flexicloud.co/blogs/rent-multi-node-b300-cluster) (2026-09-28): A B300 cluster is far more than a pile of GPUs. What to check before you commit: whether you need multi-node at all, the fabric and storage that decide performance, how reserving allocation safely works, realistic lead times, and the red flags when sourcing. - [Questions to ask a GPU provider before you sign](https://flexicloud.co/blogs/questions-to-ask-a-gpu-provider) (2026-09-24): More GPU providers than ever, and many resell capacity they do not own. The questions to ask any provider before you sign: who owns the hardware, the config, the fabric, proof it works, the datacentre, and the one almost nobody asks, what happens if they go out of business? - [NVLink, NVSwitch and InfiniBand: the fabric that makes eight GPUs act as one](https://flexicloud.co/blogs/gpu-interconnect-nvlink-infiniband) (2026-09-22): Buying GPUs is easy; making eight of them act as one is a question of fabric. How NVLink and NVSwitch scale up inside a node, how InfiniBand scales out between nodes, why a GPU that waits is money burning, and why a FlexiCloud B300 node is a single NVLink-5 domain. - [Will your model fit? A practical guide to GPU memory for AI](https://flexicloud.co/blogs/gpu-memory-guide) (2026-09-22): Whether your model fits is arithmetic, not a benchmark. The bytes-per-parameter maths for inference and training, the levers that shrink it (quantization, LoRA, checkpointing), and a back-of-envelope guide to size any job, mapped to what fits on a B300. - [Rent GPUs, buy your own, or use an API? A guide for teams building on AI](https://flexicloud.co/blogs/rent-buy-api-gpu-guide) (2026-09-22): Pay-per-token API, buy your own GPUs, or rent bare-metal nodes? The right answer is about utilisation and control, not hype. A fair look at all three, and a simple way to decide which your workload actually needs. - [What it costs to fine-tune a model on an 8-GPU B300 node](https://flexicloud.co/blogs/gpu-b300-fine-tuning-cost) (2026-09-22): The cost of a B300 job is GPU-hours × the rate, nothing more. Here is the actual arithmetic, three worked examples from a short LoRA run to a two-week training job, and an honest note on when a node is the wrong tool. - [NVIDIA B300 in India: bare-metal GPU nodes, real pricing, and who they are for](https://flexicloud.co/blogs/nvidia-b300-india-bare-metal-gpu-nodes) (2026-09-22): FlexiCloud now runs dedicated NVIDIA B300 SXM nodes in India, the EU and the US, from $4.50 per GPU-hour. What an 8-GPU node is, what it costs, bare metal vs managed, and how to tell whether you need one. - [B300 vs B200: which Blackwell node, and why the right one lowers your TCO](https://flexicloud.co/blogs/nvidia-b300-vs-b200) (2026-09-22): The B200 is 17% cheaper per GPU-hour; the B300 carries 60% more memory. Which is lower total cost is not about the sticker rate; it is about whether your workload fits in 180 GB, because memory decides how many GPUs and nodes you need. An honest B300-vs-B200 TCO guide. - [The Real Cost of Cheap Cloud Hosting: Hidden Tradeoffs](https://flexicloud.co/blogs/the-real-cost-of-cheap-cloud-hosting) (2026-01-01): The cheapest plan rarely stays cheapest. The hidden trade-offs of bargain hosting, oversold servers, slow support, surprise overage fees and downtime, and what they really cost. - [Why Are More Developers Moving to Cloud GPUs?](https://flexicloud.co/blogs/why-are-more-developers-moving-to-cloud-gpus) (2025-12-30): Developers are trading their own GPU boxes for cloud GPUs. The capex, utilisation and scaling maths behind renting instead of owning, and when each choice actually makes sense. - [FlexiCloud vs Kinsta: A Practical Comparison for High-Traffic Sites](https://flexicloud.co/blogs/flexicloud-vs-kinsta) (2025-12-27): Kinsta is a polished WordPress-only platform with automatic scaling. FlexiCloud is managed hosting on the cloud provider you choose, for any stack, with scaling you control. Here is how to tell which one fits. - [E-commerce Hosting Architecture: Solutions for High Traffic and Growth](https://flexicloud.co/blogs/e-commerce-hosting-architecture-solutions-for-high-traffic-and-growth) (2025-12-20): A shop is not a brochure site: carts cannot be cached and traffic arrives in spikes. The hosting architecture e-commerce needs, caching layers, scaling, and staying up on sale day. - [Types of Web Hosting and Which One Is Best for You](https://flexicloud.co/blogs/types-of-web-hosting-and-which-one-is-best-for-you) (2025-12-19): Website hosting is a service you’ve probably heard of countless times. Hosting your website means storing your site’s data on a server connected to the - [How to Fix the 502 Bad Gateway Error in WordPress](https://flexicloud.co/blogs/how-to-fix-the-502-bad-gateway-error-in-wordpress) (2025-12-19): A 502 bad gateway error is triggered when your Hosting for WordPress server gets an invalid response for the requested page. - [How High-Traffic Websites Should Plan Their Cloud Infrastructure?](https://flexicloud.co/blogs/how-high-traffic-sites-should-plan-their-cloud-infrastructure) (2025-12-19): High traffic breaks setups that worked at small scale. How to plan cloud infrastructure for spikes, caching, load balancing, database tuning and real headroom to grow. - [Common SSL Certificate Errors and Easy Ways to Fix Them](https://flexicloud.co/blogs/common-ssl-certificate-errors-and-easy-ways-to-fix-them) (2025-12-19): Why does your browser warn you when visiting a site with “SSL certificate problem” or “SSL connection error”? These messages are more than just annoyances. - [Best managed WordPress hosting in India: how to choose in 2026](https://flexicloud.co/blogs/best-managed-wordpress-hosting-india) (2025-12-19): What makes WordPress hosting fast and reliable for Indian visitors, what managed hosting should include, shared hosting or a managed cloud server, what it costs, and the questions to ask before you move. - [How Data Center Choice Impacts Real-World Latency?](https://flexicloud.co/blogs/how-data-center-choice-impacts-real-world-latency) (2025-12-17): Where your server sits decides how fast your site feels. How data-centre location drives real-world latency, and how to pick a region close to the users who matter most. - [How to Select the Right Cloud Server Size for Your Website?](https://flexicloud.co/blogs/how-to-select-the-right-cloud-server-size-for-your-website) (2025-12-08): Too small and your site crawls; too big and you overpay. How to size CPU, RAM and disk to your real traffic, with a sensible starting point and clear signs it is time to scale. - [How to Migrate from Shared Hosting to Cloud Without Downtime: Step-by-Step Guide](https://flexicloud.co/blogs/how-to-migrate-from-shared-hosting-to-cloud) (2025-12-06): A step-by-step guide to moving from shared hosting to a cloud server with zero downtime: backups, staging, DNS cutover, and the checks to run before and after the switch. - [GPU Cloud Computing in India: Use Cases](https://flexicloud.co/blogs/gpu-cloud-computing-in-india-use-cases) (2025-12-05): What Indian teams actually use cloud GPUs for, AI training and inference, rendering, simulation and research, and why running the compute onshore increasingly matters. - [The 3 Cloud Deployment Models: Public, Private, and Hybrid (Features, Differences, and Use cases)](https://flexicloud.co/blogs/cloud-deployment-models) (2025-12-05): Public, private or hybrid cloud? The three deployment models compared on control, cost and compliance, and the use cases each one actually fits best. - [Best Cloud Service Providers in 2026: A Complete Comparison Guide](https://flexicloud.co/blogs/best-cloud-service-providers) (2025-11-29): What a cloud service provider is, how the main ones compare in 2026, and how to pick one for your site. - [What is WordPress Caching: Types, Benefits, and How it Works](https://flexicloud.co/blogs/what-is-wordpress-caching-types-benefits-and-how-it-works) (2025-11-11): You must have observed how some WordPress sites load quickly while others take a few extra seconds.  That added slowness often comes down to WordPress caching. - [DigitalOcean vs Vultr vs Linode vs Hetzner: Which Cloud is Best for Your Needs?](https://flexicloud.co/blogs/digitalocean-vs-vultr-vs-linode-vs-hetzner-which-cloud-is-best-for-your-needs) (2025-11-11): You stare at four cloud providers (DigitalOcean vs Vultr vs Linode vs Hetzner) that offer near-identical specs: so how do you pick? The simplicity of - [What is managed cloud hosting? What you get, what it costs and when you need it](https://flexicloud.co/blogs/what-is-managed-cloud-hosting) (2025-11-01): Managed cloud hosting gives you your own cloud server and a team that runs it: setup, security patches, monitoring, backups and support. How it differs from unmanaged cloud, shared hosting and a VPS, what it costs, and how to choose a provider. - [The Unified Hosting Dashboard: Multi Cloud Management, Billing, and Support in One Place](https://flexicloud.co/blogs/unified-hosting-dashboard-multi-cloud-management-billing-and-support-in-one-place) (2025-11-01): What if there's only one place to do everything related to your website? Like, pick hosting plans, managing files, billing, and replying to support tickets? - [Understanding Nginx: What is Nginx and How Does Nginx Work?](https://flexicloud.co/blogs/understanding-nginx-what-is-nginx-and-how-does-nginx-work) (2025-11-01): NGINX is open-source web server software utilized for reverse proxy, load balancing, and caching. It offers HTTPS server capabilities and is primarily engineered for optimal performance and stability. Additionally, it serves as a proxy server for email communication protocols like IMAP, POP3, and SM - [Top 9 cPanel Alternatives for Better Performance, Freedom, and Value](https://flexicloud.co/blogs/top-9-cpanel-alternatives-for-better-performance-freedom-and-value) (2025-11-01): cPanel is no longer the only option, and its licensing keeps climbing. Nine alternatives, Webuzo, Plesk, CyberPanel and more, compared on performance, features and price. - [How Web Hosting Affects Your SEO Rankings (and How to Choose the Right One)](https://flexicloud.co/blogs/how-web-hosting-affects-your-seo-rankings) (2025-11-01): Does web hosting affect SEO more than you think? This question always arises in the minds of website owners nowadays. The answer is very simple- a reliable, - [How to Prevent Downtime During Hosting Migrations](https://flexicloud.co/blogs/how-to-prevent-downtime-during-hosting-migrations) (2025-11-01): Move hosts without your site going dark. The migration playbook, test on a temporary URL, sync data, lower DNS TTL, and cut over only once every check passes. - [How to Install WordPress Without cPanel (Using FlexiCloud Panel)?](https://flexicloud.co/blogs/how-to-install-wordpress-without-cpanel) (2025-11-01): You do not need cPanel to run WordPress. Install it with the FlexiCloud Panel, database, files and SSL, in a few clicks, no command line and no cPanel licence required. - [A Complete Guide to Redis Installation on Ubuntu 22.04](https://flexicloud.co/blogs/how-to-install-and-configure-redis-on-ubuntu-22-04) (2025-11-01): If you’ve ever wanted your apps to be faster, Redis is your friend. Short for Remote Dictionary Server, Redis stores data in memory, which means your app can - [How to Find and Change Your WordPress Login URL in Different Ways: The Ultimate Guide](https://flexicloud.co/blogs/how-to-find-and-change-your-wordpress-login-url-in-different-ways) (2025-11-01): Lost your wp-admin URL, or want to change it to cut down on bot attacks? Every way to find and change your WordPress login URL, and the security reasons to bother. - [Cloud hosting vs VPS hosting: Understanding the Features, Benefits, and Key Differences](https://flexicloud.co/blogs/cloud-hosting-vs-vps-hosting) (2025-11-01): It's your first day of launching your website, and you are waiting for visitors, but it keeps crashing. Has this stressful event happened to you? That’s - [What is Linode](https://flexicloud.co/blogs/what-is-linode) (2025-09-23): This blog will explain Linode in detail, starting with its definition and operation. We'll look at the benefits of using Linode for development projects, such - [What is Vultr](https://flexicloud.co/blogs/what-is-vultr) (2025-09-15): Today, cloud computing is a vital part of many industries, like retail, travel, e-commerce, telecommunications, banking, and pharmaceuticals. These industries - [What is Hetzner](https://flexicloud.co/blogs/what-is-hetzner) (2025-09-15): Hetzner is a German company that provides web hosting services and data centers. They offer a range of services, from cloud computing and dedicated servers to - [What is DigitalOcean](https://flexicloud.co/blogs/what-is-digitalocean) (2025-09-15): Platforms for Infrastructure as a Service (IaaS) designed specifically for software developers are offered by cloud computing business DigitalOcean. It faces - [How to Fix the Mixed Content Error in WordPress?](https://flexicloud.co/blogs/how-to-fix-the-mixed-content-error-in-wordpress) (2025-09-15): When a user accesses a web page via HTTPS, their connection with the web server is secured through TLS, providing protection against most sniffers and man-in-the-middle attacks. However, if an HTTPS page contains content fetched using unencrypted HTTP, it is referred to as a mixed content page. Such - [How to convert a WordPress site from HTTP to HTTPS](https://flexicloud.co/blogs/how-to-convert-wordpress-site-from-https-to-https) (2025-09-15): HTTP stands for Hypertext Transfer Protocol, but when it comes to securing your WordPress website, HTTPS (Hypertext Transfer Protocol Secure) is the preferred - [How to Install Django on Ubuntu](https://flexicloud.co/blogs/how-to-install-django-on-ubuntu) (2025-09-03): Django is a popular and open-source web framework built with Python that allows developers to build web applications quickly and efficiently. It emphasizes - [How to install and set up PostgreSQL](https://flexicloud.co/blogs/postgresql-installation) (2025-08-29): Having been actively developed for over three decades, PostgreSQL is a robust and sophisticated object-relational database management system that is - [BUNNY CDN](https://flexicloud.co/blogs/bunny-cdn) (2025-08-29): A Content Delivery Network is a network of servers across a diverse selection of geographic locations that help improve the speed and quality of content ## Research ### India Data Centre Market Guide 2026 URL: https://flexicloud.co/india-data-centers Last modified: 2026-10-24 India's data centre market: 1.6 to 1.8 GW live in mid-2026, 75 projects tracked in October 2026, and guides to operators, power, cables and entering India. India had about 1.6 to 1.8 GW of data centre capacity live in mid-2026, about half of it in Mumbai (JLL; Cushman & Wakefield). This guide tracks 75 projects from 31 developers that are building or planning more, and links to seven topic guides, from the main operators to power costs and how a GPU cloud can enter India. Every figure is sourced and dated. We keep built capacity and announcements apart. Most of the money pledged for India is still a plan. Topic pages: - [Who runs India's data centres, and how much is live?](https://flexicloud.co/india-data-centers/operators): Live MW and vacancy for seven cities, led by Mumbai with 890 MW (H1 2026), plus live capacity, pipeline and owners for 14 major operators, each figure with its date and source. - [Is India a good data centre market, and what are the risks?](https://flexicloud.co/india-data-centers/why-india): India produces nearly 20% of the world's data but hosts about 3% of its data centres (Jun 2025). See the capacity forecasts to 2030, the demand behind them, and the risks: grid delays, water, unsettled rules, vacancy and unbuilt announcements. - [How much are AWS, Microsoft and Google investing in India, and why?](https://flexicloud.co/india-data-centers/hyperscaler-investment): Amazon, Microsoft and Google have pledged more than US$56bn for AI and cloud in India, mostly for 2026 to 2030 (Oct 2026). See each pledge, what is already live, who else is investing and why. - [How do data centre power, build and colocation costs in India compare with other countries?](https://flexicloud.co/india-data-centers/power-costs): Grid power in India is mid-priced, building is cheap and colocation rent is low by global standards. See FY2026-27 state power tariffs, build cost per MW and city colocation rents, set against Singapore, Europe, the US and other Asian markets. - [Which subsea cables land in India, and what is coming?](https://flexicloud.co/india-data-centers/subsea-cables): About 15 international cables are live in India, mainly landing in Mumbai and Chennai, and 5 more are planned (Oct 2026). See every live and upcoming cable, landing stations by city, repair times and what it means for choosing a data centre site. - [How are India's data centres connected?](https://flexicloud.co/india-data-centers/connectivity): Where India's networks meet, what an internet exchange port costs, which buildings host AWS, Azure, Google Cloud and Oracle on-ramps, and typical latency between cities. Mumbai leads: DE-CIX had 451 networks there on 14 Oct 2026. - [How can a GPU cloud (neocloud) enter India?](https://flexicloud.co/india-data-centers/neocloud-guide): Entry models from colocation to build-to-suit, the Indian and US rules that apply (CERT-In, DPDP, export controls), IndiaAI reference GPU prices, verified state incentives and a 14-point checklist. Also why the 2047 tax exemption probably does not fit a neocloud that owns its GPUs. **How big is India's data centre market in 2026?** About 1.6 to 1.8 GW of data centre capacity was live in mid-2026. JLL puts it at 1.6 GW (Aug 2026) and Cushman & Wakefield at 1,789 MW (H1 2026), with Mumbai holding about half. Cushman & Wakefield also counts 509 MW under construction and 3,351 MW planned to 2030 (H1 2026). AWS, Microsoft and Google have pledged more than US$56bn for AI and cloud in India, mostly for 2026 to 2030 (Oct 2026), but pledges are not built capacity. **Where are the biggest data centre projects in India?** The largest plans are six gigawatt-scale announcements: AirTrunk at Raigad near Mumbai (3 GW, a letter of intent for land), two new 2 GW platforms with no sites disclosed yet, TeraNova DC (Hillhouse, over 2 GW across five states by 2032) and Cortex DC (The Chatterjee Group, 2 GW over about seven years, per PTI), and about 1 GW each from AM Group at Jewar, DigitalConnexion at Visakhapatnam and TCS HyperVault at Hyderabad (tracker, 24 October 2026). Google is building a gigawatt-scale campus at Visakhapatnam (ground broken 28 Apr 2026) but publishes no phase MW. For live capacity, Mumbai, including Navi Mumbai, leads with 890 MW, and Hyderabad has the second-largest pipeline at 710 MW (Cushman & Wakefield, H1 2026). **How fast is India's data centre capacity growing?** MeitY says capacity rose from about 375 MW in 2020 to about 1,500 MW by 2025 (PIB, Mar 2026). Cushman & Wakefield counted 1,280 MW live in H1 2025 and 1,789 MW in H1 2026, though part of that rise is a change in what it counts in Bengaluru. CBRE expects about 2 GW by end-2026 (Oct 2026). Most forecasts for 2029 to 2030 fall between 4 and 6 GW; the Why India page of this guide compares them. **How do I use the India data centre projects tracker?** The tracker lists 75 projects from 31 developers that are under construction, expanding or announced, each re-checked in October 2026. It keeps two totals apart: about 2.5 GW under construction or development in 38 projects, and about 15.0 GW announced in 37 projects, 10.0 GW of which is 6 gigawatt-scale plans (24 October 2026). MW is the figure each developer gives, IT load or total power, and Not disclosed means we found no reliable figure. Use the filters or search to find a city, developer or status. These totals are not comparable with Cushman & Wakefield's 509 MW, which counts only IT load being built in the top 7 cities. **We are a neocloud looking at India. Where does FlexiCloud fit in?** In two ways. If you need GPUs now, we source dedicated NVIDIA nodes, such as B200 and B300, through more than 10 partner providers in India, the EU and the US. If you are planning your own capacity, our DC advisory helps with site and partner selection, power and connectivity, the compliance path, commercial terms and introductions to Indian customers. We do not own data centres, and if any party we recommend will pay us, we tell you in writing before you sign. Start by booking a call. Projects tracker (75 projects, re-checked October 2026; MW as given by the developer, IT load or total power): - Google, Visakhapatnam AI hub (gigawatt-scale campus) (Tarluvada, Adavivaram and Rambilli, Visakhapatnam / Anakapalli districts, Andhra Pradesh): Not disclosed. Under construction or development: Under construction (ground broken 28 Apr 2026); no phase MW published; about 1 GW reported for the full campus, Google says gigawatt-scale. Commissioning: 2026 to 2030. Backer: Google (via Raiden Infotech); AdaniConneX and Nxtra by Airtel as partners. - Google, Navi Mumbai leased DC (EverYondr Mumbai 1) (MIDC, Navi Mumbai): 60 MW. Under construction or development: Under construction; 60 MW total and 30 MW Phase 1 reported, not confirmed; completion slipped from 2025. Commissioning: 2026 to 2027. Backer: Google as tenant (Raiden Infotech, 28-year lease); built by EverYondr. - AdaniConneX, Noida campus (Sector 62, Noida): 50 MW. Under construction or development: Partly operational; 50 MW is the whole campus, including live halls; live MW not disclosed. Commissioning: 2025 to 2027. Backer: Adani Enterprises + EdgeConneX (JV). - AdaniConneX, Taloja AI campus (Taloja, Navi Mumbai, Maharashtra): 240 MW. Announced (MoU, land, plan): Announced Oct 2026: 3 x 80 MW buildings in design; option for 240 MW more. Commissioning: 2028+ (first phase mid-2028). Backer: Adani Enterprises + EdgeConneX; a US AI cloud company is sole tenant. - AdaniConneX, Chennai 1, Building 2 (SIPCOT IT Park, Siruseri, Chennai): 16 MW. Announced (MoU, land, plan): Planned; Building 1 (17 MW IT) already live; campus 33 MW IT. Commissioning: 2027 (target). Backer: Adani Enterprises + EdgeConneX. - AdaniConneX, Hyderabad campus (Elkatta) (Elkatta, near Shadnagar, Rangareddy district, Telangana): Not disclosed. Under construction or development: Partly operational (2025); further phases planned; MW not disclosed. Commissioning: 2025 to 2028. Backer: Adani Enterprises + EdgeConneX. - Reliance Industries, Jamnagar AI data centre, first phase (Jamnagar, Gujarat): 120 MW. Under construction or development: Under construction; more than 120 MW due H2 2026, not yet announced live; multi-GW planned (NVIDIA, 2023: eventually 2,000 MW); separate 168 MW built-to-suit DC for Meta, option to scale (RIL, Jun 2026). Commissioning: H2 2026 (first phase). Backer: Reliance Industries / Reliance Intelligence. - Reliance Industries, Vizianagaram AI data centre + cable landing station (Polipalli and Bhogapuram, Vizianagaram district, Andhra Pradesh): Not disclosed. Announced (MoU, land, plan): Land allotted by state order, 20 May 2026; 854.97 acres; ₹1.08 lakh crore; MW not disclosed (press: about 1.5 GW). Commissioning: Not announced. Backer: Reliance Industries. - CtrlS Datacenters, Chandan Valley campus (Chandanvelly, near Hyderabad, Telangana): 612 MW. Announced (MoU, land, plan): Announced Jan 2025; Phase 1 power (250 MW) sanctioned; ICRA (Oct 2026) lists 2 CtrlS DCs under construction in Hyderabad without naming the sites. Commissioning: Not announced. Backer: CtrlS (founder-led) + CPP Investments. - CtrlS Datacenters, Mumbai DC6 & DC7 (Mahape campus) (TTC MIDC Mahape, Navi Mumbai): 95 MW. Under construction or development: Under construction (DC6 & DC7); DC1 to DC5 live. Commissioning: 2026 to 2027. Backer: CtrlS (founder-led) + CPP Investments. - CtrlS Datacenters, Kolkata campus, Phases 2 to 4 (Bengal Silicon Valley, New Town, Kolkata): 44 MW. Announced (MoU, land, plan): Planned; Phase 1 (16 MW IT) live since Aug 2025; campus 60+ MW IT. Commissioning: Not announced. Backer: CtrlS. - CtrlS Datacenters, Bhopal edge data centre (Badwai IT Park, Bhopal): 12 MW. Under construction or development: Ground broken Apr 2025 (phased). Commissioning: 2026 to 2027 (estimate). Backer: CtrlS. - Yotta Data Services, NM2 (Panvel park) (Hiranandani Fortune City, Panvel, Navi Mumbai): 80 MW. Under construction or development: Fit-out under way (core and shell ready); 75 to 80 MW reported; NM1 (52 MW IT) live. Commissioning: 2026 to 2027. Backer: Hiranandani Group. - Yotta Data Services, Greater Noida park: D2 + D4 (Greater Noida, Uttar Pradesh): 180 MW. Under construction or development: D2 (60 MW) due 2026, not yet announced live; checked 14 Oct 2026; D4 (120 MW) under construction (reported). Commissioning: 2026 to 2027. Backer: Hiranandani Group. - NTT Data GDC, NAV2 Navi Mumbai campus (remaining buildings) (MIDC Road, Navi Mumbai): Not disclosed. Under construction or development: Partly live; campus designed for 336 MW IT (500 MW power); unbuilt MW not disclosed. Commissioning: 2025 to 2027. Backer: NTT DATA (Japan); Tokyo Century co-investor. - NTT Data GDC, Noida 2, Building B (Plot 18, Noida data centre corridor): 30.4 MW. Under construction or development: May already be live (Baxtel); NTT has not announced it; Building A (22.4 MW IT) live since Jan 2024. Commissioning: 2025 to 2026. Backer: NTT DATA (Japan). - NTT Data GDC, Chennai 2, Building B (Ambattur, Chennai): 17.4 MW. Under construction or development: Under construction; Building A (17.4 MW IT) live since 2023. Commissioning: 2027 (estimate). Backer: NTT DATA (Japan). - STT GDC India, Navi Mumbai 2 & 3 (MIDC Mahape, Navi Mumbai): 91 MW. Under construction or development: Under development (ground broken Nov 2024); Navi Mumbai 1 (28 MW IT) appears live. Commissioning: 2026 to 2027. Backer: STT GDC (KKR 75% + Singtel 25% since Sep 2026). - STT GDC India, Navi Mumbai 4 (MIDC Mahape, Navi Mumbai (per STT)): 36 MW. Under construction or development: Under development per STT website; no dated announcement. Commissioning: Not announced. Backer: STT GDC (KKR 75% + Singtel 25% since Sep 2026). - STT GDC India, Palava campus, Phase 1 (Palava, Thane district (MMR), Maharashtra): 50 MW. Under construction or development: Ground broken Mar 2026 (50 MW IT); STT lists 171 MW under development; campus designed for about 400 MW. Commissioning: Not announced. Backer: STT GDC (KKR 75% + Singtel 25% since Sep 2026). - STT GDC India, Devanahalli campus (KIADB land) (Devanahalli, Bengaluru): 100 MW. Announced (MoU, land, plan): Announced by Karnataka Govt, 25 Sep 2026 (about ₹5,000 crore); no MoU or timeline reported. Commissioning: Not announced. Backer: STT GDC (KKR 75% + Singtel 25% since Sep 2026). - STT GDC India, Noida 2 & 3 (Sector 143A, Noida): 50 MW. Under construction or development: Under development; Noida 1 (17 MW IT) live. Commissioning: Not announced. Backer: STT GDC (KKR 75% + Singtel 25% since Sep 2026). - STT GDC India, Chennai 4 (Ambattur campus) (Ambattur, Chennai): 18 MW. Under construction or development: Ground broken Feb 2026; 40 MW already live on this campus; campus potential about 130 MW. Commissioning: Not announced. Backer: STT GDC (KKR 75% + Singtel 25% since Sep 2026). - STT GDC India, Siruseri campus build-out (beyond Chennai 7) (SIPCOT IT Park, Siruseri, Chennai): 37.8 MW. Announced (MoU, land, plan): Planned; Chennai 7 Phase 1 (7.2 MW) live Feb 2026; campus designed for 45 MW. Commissioning: Not announced. Backer: STT GDC (KKR 75% + Singtel 25% since Sep 2026). - STT GDC India, Hyderabad AI-ready campus (Meerkhanpet, Mucherla, Ranga Reddy district, Telangana): 100 MW. Announced (MoU, land, plan): MoU with Telangana Govt (Jan 2025); up to ₹3,500 crore. Commissioning: Not announced. Backer: STT GDC (KKR 75% + Singtel 25% since Sep 2026). - Nxtra by Airtel, Mahape hyperscale park (MUM 4) (TTC Industrial Area, Mahape, Navi Mumbai): 42 MW. Under construction or development: Phased; first tower built 2025 (Baxtel); Nxtra lists MUM 3 at 18 MW IT and MUM 4 at 42 MW IT without saying which is live. Commissioning: 2026 to 2027. Backer: Bharti Airtel (control); Alpha Wave, Carlyle, Anchorage (US$1bn round agreed Mar 2026). - Nxtra by Airtel, New AI-ready Chennai campus (Chennai (site not disclosed)): Not disclosed. Announced (MoU, land, plan): In development per Airtel (Mar 2026); site and MW not disclosed. Commissioning: Not announced. Backer: Bharti Airtel (control); Alpha Wave, Carlyle, Anchorage (US$1bn round agreed Mar 2026). - Nxtra by Airtel, Kolkata hyperscale DC (Action Area IIE, New Town (Bengal Silicon Valley), Kolkata): 25 MW. Under construction or development: Under construction since Nov 2022; operations expected about Sep 2026; no launch announced by 14 Oct 2026. Commissioning: 2026. Backer: Bharti Airtel (control); Alpha Wave, Carlyle, Anchorage (US$1bn round agreed Mar 2026). - Sify Technologies, Rabale campus, Towers 6 to 12 (Rabale, Navi Mumbai): 276 MW. Under construction or development: T6/T7 and T11/T12 under construction or advanced development; T1 to T5 built (101.26 MW IT); 377+ MW IT design. Commissioning: Not announced (phased). Backer: Sify Infinit Spaces (Sify Technologies, Nasdaq: SIFY); IFC loan Jun 2026. - Sify Technologies, Hyderabad 02 campus (Fab City, Hyderabad): 250 MW. Announced (MoU, land, plan): Planned (250+ MW IT design); not in Sify list of DCs under development (Jul 2026). Commissioning: Not announced. Backer: Sify Infinit Spaces. - Sify Technologies, Noida 02 campus, remaining towers (B-11/12/13, Sector 132, Noida): 104 MW. Under construction or development: 2 more towers to come (130+ MW IT design); Tower B live (25.92 MW IT built). Commissioning: Not announced. Backer: Sify Infinit Spaces. - Sify Technologies, Chennai 02 campus, remaining towers (SIPCOT IT Park, Siruseri, Chennai): 104 MW. Under construction or development: More towers to come (130+ MW IT design); Tower B live (25.92 MW IT built). Commissioning: Not announced. Backer: Sify Infinit Spaces; IFC loan Jun 2026. - Sify Technologies, Bengaluru 02 campus (KIADB Aerospace Park, Bengaluru): 40 MW. Under construction or development: Under development per Sify (FY26 integrated report); no date disclosed. Commissioning: Not announced. Backer: Sify Infinit Spaces. - Sify Technologies, Vizag AI edge DC + open cable landing station (Rushikonda-Madhurawada IT Park, Visakhapatnam): 50 MW. Under construction or development: Foundation stone laid Oct 2025 (₹1,500 crore, two phases). Commissioning: Not announced. Backer: Sify Infinit Spaces. - Sify Technologies, Vizag hyperscale campus (Paradesipalem) (Paradesipalem, Visakhapatnam): 500 MW. Announced (MoU, land, plan): State-approved (SIPB Jul 2025; about ₹15,266 crore); 12.4 of 25 extra acres allotted conditionally (Jun 2026). Commissioning: Not announced. Backer: Sify Infinit Spaces. - CapitaLand (CLINT), CapitaLand DC Mumbai, Tower 2 (Airoli, Navi Mumbai): 37 MW. Under construction or development: Completion 4Q 2026, handover 1H 2027; pre-leased to a hyperscaler; Tower 1 (34 MW IT) live. Commissioning: 2026 to 2027. Backer: CapitaLand India Trust (79.8%) + CapitaLand India Data Centre Fund (20.2%). - CapitaLand (CLINT), CapitaLand DC Chennai (Ambattur) (Ambattur, Chennai): 34 MW. Under construction or development: Core and shell works in progress; 34 MW IT (53 MW power); no tenant signed (Jul 2026). Commissioning: 4Q 2026 (target). Backer: CapitaLand India Trust (79.8%) + CapitaLand India Data Centre Fund (20.2%). - CapitaLand (CLINT), CapitaLand DC ITPH (International Tech Park Hyderabad, Madhapur): 27 MW. Under construction or development: Phase 1 nearing completion (Jul 2026); core and shell plus Phase 1 M&E; completion not announced by 14 Oct 2026; 27 MW IT (42 MW power). Commissioning: 2026 (3Q target passed). Backer: CapitaLand India Trust (79.8%) + CapitaLand India Data Centre Fund (20.2%). - Colt DCS, Navi Mumbai campus, later phases (Airoli, Navi Mumbai): 112 MW. Announced (MoU, land, plan): Later phases planned, no construction date disclosed; Phase 1 (22 MW IT) live since Sep 2023. Commissioning: Not announced. Backer: Colt DCS (Fidelity) + RMZ, 50:50 JV. - Colt DCS, Chennai campus (Ambattur, Chennai): 96 MW. Announced (MoU, land, plan): Planned (10-acre site bought Mar 2024). Commissioning: 2027 (target). Backer: Colt DCS (Fidelity) + RMZ, 50:50 JV. - Princeton Digital Group (PDG), MU1 Navi Mumbai expansion (Airoli, Navi Mumbai): 100 MW. Under construction or development: First phase of a 100 MW expansion under construction; 50 MW live in 2 buildings. Commissioning: 2026 (first phase). Backer: Warburg Pincus (largest), Stonepeak, OTPP, Mubadala. - Princeton Digital Group (PDG), MU2 Navi Mumbai (Navi Mumbai, Maharashtra): 120 MW. Under construction or development: Construction started Apr 2026 (single 5-storey building). Commissioning: Not announced. Backer: Warburg Pincus (largest), Stonepeak, OTPP, Mubadala. - Princeton Digital Group (PDG), HY1 Hyderabad (Hyderabad, Telangana): 90 MW. Announced (MoU, land, plan): Site acquisition announced Mar 2026; power and key land approvals in place. Commissioning: Not announced. Backer: Warburg Pincus (largest), Stonepeak, OTPP, Mubadala. - Princeton Digital Group (PDG), CH1 Chennai (North Chennai Metropolitan Area, Tamil Nadu): 72 MW. Announced (MoU, land, plan): Announced Sep 2024 (9.3 acres); no construction start confirmed. Commissioning: Not announced. Backer: Warburg Pincus (largest), Stonepeak, OTPP, Mubadala. - Iron Mountain Data Centers (ex-Web Werks), MUM-3 AI hyperscale campus (Mumbai area (2023 plan named Rabale, Navi Mumbai)): 85 MW. Under construction or development: Ground broken Jan 2026 (reported); 85 MW. Commissioning: 2027 to Q1 2028. Backer: Iron Mountain (100% owner). - Iron Mountain Data Centers (ex-Web Werks), CHN-1 & CHN-2 (Ambattur campus) (Ambattur, Chennai): 42 MW. Under construction or development: CHN-1 (23.2 MW) topped out Apr 2026; go-live expected 2026, not announced by 14 Oct 2026; CHN-2 (19 MW) from 2028; 42 MW power. Commissioning: 2026 to 2028. Backer: Iron Mountain (100% owner). - DigitalConnexion, Chennai campus, later buildings (Ambattur Industrial Estate, Chennai): 80 MW. Announced (MoU, land, plan): Later buildings planned, no dates disclosed; MAA10 phase 1 (20 MW) live since Jan 2024; 100 MW, 4-building campus. Commissioning: Not announced. Backer: Brookfield + Reliance Industries + Digital Realty (JV). - DigitalConnexion, BOM10 Mumbai (Chandivali, Mumbai): 35 MW. Announced (MoU, land, plan): Operator lists Planned - 2026; 35 MW IT (40 MW power); no construction milestone found. Commissioning: 2026 to 2027. Backer: Brookfield + Reliance Industries + Digital Realty (JV). - DigitalConnexion, Vizag AI-native campus (Visakhapatnam, Andhra Pradesh): 1 GW. Announced (MoU, land, plan): MoU with AP Govt (Nov 2025); about US$11bn by 2030; 400 acres. Commissioning: 2030 (target). Backer: Brookfield + Reliance Industries + Digital Realty (JV). - Microsoft, Pune region expansion (Pune (Central India region)): Not disclosed. Announced (MoU, land, plan): Expansion announced Dec 2025; capacity and sites not disclosed. Commissioning: 2026 to 2029. Backer: Microsoft. - Amazon Web Services (AWS), Hyderabad region expansion (Bharat Future City) (Bharat Future City, Ranga Reddy district, Telangana): Not disclosed. Under construction or development: Ground broken 15 Jul 2026; MW not disclosed. Commissioning: 2026 to 2030 (investment period). Backer: Amazon. - Amazon Web Services (AWS), Taloja campus (Asode and Burdul) (Taloja, Navi Mumbai, Maharashtra): 473 MW. Announced (MoU, land, plan): Environmental clearance applied Apr 2026; 473 MW is a power requirement, not IT load. Commissioning: 2027 to 2030 (estimate). Backer: Amazon. - AirTrunk (Lumina CloudInfra), BOM1 Navi Mumbai (ex-Lumina) (Navi Mumbai, Maharashtra): 62 MW. Under construction or development: About 60 MW reported ready (Jan 2026 press); 62+ MW campus; AirTrunk lists the site but no launch announced by 14 Oct 2026. Commissioning: 2026. Backer: Blackstone + CPP Investments (AirTrunk owners). - AirTrunk (Lumina CloudInfra), Second Navi Mumbai site (Mahape) (Mahape, Navi Mumbai): 120 MW. Announced (MoU, land, plan): Under development (Jan 2026 press); press puts the two Navi Mumbai sites at 180 MW in total. Commissioning: Not announced. Backer: Blackstone + CPP Investments (AirTrunk owners). - AirTrunk (Lumina CloudInfra), Chandivali data centres (2 buildings) (Chandivali, Mumbai): 60 MW. Announced (MoU, land, plan): Land acquired; two 30 MW DCs planned (60 to 90 MW reported). Commissioning: Not announced. Backer: Blackstone + CPP Investments (AirTrunk owners). - AirTrunk (Lumina CloudInfra), MAA1 Chennai (Ambattur) (Ambattur, Chennai): 216 MW. Announced (MoU, land, plan): Land bought (16 acres, Jan 2026); 216 MW initial IT load; no construction start reported. Commissioning: Not announced. Backer: Blackstone + CPP Investments (AirTrunk owners). - AirTrunk (Lumina CloudInfra), HYD1 Hyderabad (Hyderabad, Telangana): 250 MW. Announced (MoU, land, plan): Under development (Jan 2026 press: two sites, about 250 MW); AirTrunk gives no MW. Commissioning: Not announced. Backer: Blackstone + CPP Investments (AirTrunk owners). - AirTrunk (Lumina CloudInfra), Raigad AI campus (Pen Growth Centre) (Pen, Raigad district (MMR), Maharashtra): 3 GW. Announced (MoU, land, plan): Letter of intent for land (Maharashtra, 1 Jun 2026); about ₹2 lakh crore reported. Commissioning: Not announced (AirTrunk has a 5 GW India plan). Backer: Blackstone + CPP Investments (AirTrunk owners). - L&T-Vyoma (ex-Cloudfiniti), Mahape campus, Phase 1 (Mahape, Navi Mumbai): 40 MW. Under construction or development: Ground broken 21 Jan 2026; first step to a 100 MW campus. Commissioning: Not disclosed. Backer: Larsen & Toubro (DC business moving to Vyoma.AI). - L&T-Vyoma (ex-Cloudfiniti), Chennai AI campus expansion (Chennai (Kanchipuram district)): 220 MW. Announced (MoU, land, plan): Design and feasibility only (L&T filing, Aug 2026); Phase 1 designed for 250 MW with 150 MVA ready; 30 MW already live. Commissioning: Not announced. Backer: Larsen & Toubro (DC business moving to Vyoma.AI). - Anant Raj, Manesar, Panchkula & Rai campuses (Manesar (Gurugram), Panchkula and Rai (Sonipat), Haryana): 279 MW. Under construction or development: 63 MW targeted by end-FY27 and 117 MW by FY28; 28 MW IT live; sources conflict on the remainder (279 to 329 MW). Commissioning: 2026 to 2032. Backer: Anant Raj Ltd (listed); DC arm to be demerged into Ashok Cloud. - Pi Data Centers, Mumbai data centre (Mumbai (site not disclosed)): 3 MW. Under construction or development: Launched Apr 2026; go-live scheduled Aug 2026, not announced by 14 Oct 2026. Commissioning: 2026. Backer: Pi Data Centers. - Techno Electric, Noida data centre (Noida, Uttar Pradesh): 30 MW. Under construction or development: Full 30 MW launching 2026 per company; Phase 1 (0.8 MW) live. Commissioning: 2026 to 2027. Backer: Techno Electric (listed). - Techno Electric, Kolkata data centre (Kolkata, West Bengal): 13 MW. Announced (MoU, land, plan): Planned; launching 2027 per company. Commissioning: 2027. Backer: Techno Electric (listed). - Indian GPU cloud (Blackstone-backed), Hyderabad AI cluster (with NTT DATA) (Hyderabad, Telangana): 400 MW. Announced (MoU, land, plan): Announced (MoU, Apr 2025, ₹10,500 crore); no construction reported. Commissioning: Not announced. Backer: Blackstone (majority, Feb 2026); NTT DATA + Telangana Govt MoU. - TCS HyperVault, Hyderabad AI campus (Bharat Future City) (Bharat Future City, Hyderabad, Telangana): 1 GW. Announced (MoU, land, plan): Land acquired (264 acres, Sep 2026); up to 1 GW, phased by demand. Commissioning: Not announced (phased). Backer: TCS + TPG (up to ₹18,000 crore platform, Nov 2025); OpenAI first customer (100 MW, Feb 2026). - Digital Edge, Navi Mumbai campus (BOM1 to BOM3) (Navi Mumbai, Maharashtra): 120 MW. Under construction or development: BOM2 topped out and BOM3 started (Aug 2026); close to 120 MW across BOM1 to BOM3, including live BOM1 (not split out); 400 MW campus plan. Commissioning: 2026 to 2028 (estimate). Backer: Stonepeak. - Digital Edge, Palava campus (Palava, Thane district (MMR), Maharashtra): 270 MW. Announced (MoU, land, plan): Land bought (about 30 acres, Jul 2026, about ₹1,000 crore); no timeline. Commissioning: Not announced. Backer: Stonepeak. - AM Group (AM Intelligence Labs), Jewar AI compute hub (YEIDA) (Sector 28, YEIDA, Jewar, Uttar Pradesh): 1 GW. Announced (MoU, land, plan): Letter of intent for land (YEIDA, Feb 2026); Phase I 384 MW (₹26,910 crore) targeted 2028. Commissioning: 2028 to 2030 (target). Backer: AM Group (Greenko founders). - UPC Volt, Bharat Future City AI-ready DC (Bharat Future City, Hyderabad, Telangana): 100 MW. Announced (MoU, land, plan): MoU with Telangana Govt (Davos, Jan 2026; ₹5,000 crore). Commissioning: Not announced. Backer: UPC Renewables + VOLT Data Centers (JV). - Trentar (Dorf Ketal Group), Green hyperscale DC park (Maharashtra (site not disclosed)): 500 MW. Announced (MoU, land, plan): MoU with Maharashtra Govt (Oct 2026; ₹30,000 crore); feasibility studies pending. Commissioning: Not announced. Backer: Dorf Ketal Group. - HCLTech + Sarvam AI, Odisha sovereign AI park data centre (Khurda district (near Bhubaneswar), Odisha): Not disclosed. Announced (MoU, land, plan): MoU Jul 2026 (₹14,257 crore incl. state assistance); MW not disclosed. Commissioning: Not announced. Backer: HCLTech + Sarvam AI; Govt of Odisha support. - Sarvam AI, Tamil Nadu sovereign AI park (Chennai, Tamil Nadu): Not disclosed. Announced (MoU, land, plan): MoU Jan 2026 (about ₹10,000 crore); MW not disclosed. Commissioning: Not announced. Backer: Sarvam AI + Govt of Tamil Nadu. - TeraNova DC (Hillhouse), India platform, sites not yet disclosed (Maharashtra, Telangana, Andhra Pradesh, Delhi NCR and Gujarat): 2 GW. Announced (MoU, land, plan): Platform launched 15 Oct 2026: over 2 GW and about US$12 to 15bn by 2032; no sites, MW per site or first go-live disclosed yet. Commissioning: By 2032. Backer: Hillhouse Investment (Singapore); CEO Sanjay Bhutani. - Cortex DC (The Chatterjee Group), India AI-ready platform, sites not yet disclosed (Not disclosed (the company says its capacity is anchored on the Golden Quadrilateral corridor)): 2 GW. Announced (MoU, land, plan): Platform launched 22 Oct 2026: 2 GW (IT or total power not stated); first facility targeted within 18 months of first shovel, no start date yet; 51% of energy from green sources; no sites, investment or MW per site disclosed. Commissioning: About 7 years (per PTI). Backer: The Chatterjee Group (TCG, New York); CEO Rahul Sharma. ### Who runs India's data centres, and how much is live? URL: https://flexicloud.co/india-data-centers/operators Last modified: 2026-10-23 Who runs India's data centres: Mumbai held 890 MW live in H1 2026, NTT DATA 237 MW and Sify 134 MW. Live MW, vacancy and pipeline by city and operator. Mumbai held 890 MW of live data centre capacity in H1 2026, about half of India's total. Among operators, NTT DATA had 237 MW live (IFC, June 2026) and Sify 134 MW (July 2026). Operators count MW in different ways and some campuses have slipped, so get the energisation date in writing. Key facts: - Mumbai, including Navi Mumbai, had 890 MW live, about half of the 1,789 MW live in India's top 7 cities. Its colocation vacancy was 7.9%. (as of H1 2026; Cushman & Wakefield) - After Mumbai came Chennai (214 MW), Delhi NCR (183 MW), Bengaluru (162 MW), Hyderabad (151 MW) and Pune (133 MW) of live IT load, the power available to servers. (as of H1 2026; Cushman & Wakefield) - NTT DATA GDC India ran 237 MW in 16 data centres, with 105 MW under construction in 5 data centres. (as of 22 Jun 2026; IFC) - Sify Infinit Spaces had 134 MW live and earning revenue, and 188 MW built. (as of 15 Jul 2026; Sify Technologies, Q1 FY2027 earnings call) - CtrlS had 150 MW operational in FY2026; ICRA expects 950 to 1,000 MW by FY2029. (as of 5 Oct 2026; ICRA) **Who are the largest data centre operators in India?** The biggest are Indian groups such as Nxtra by Airtel, CtrlS and Sify, and Asian and global players such as STT GDC India and NTT DATA. On published figures, NTT DATA had 237 MW live in 16 data centres (June 2026), Nxtra about 250 MW of power capacity (July 2026), CtrlS 150 MW operational per ICRA (October 2026) and Sify 134 MW live (July 2026). STT GDC India reports 400+ MW of critical IT across 30 facilities (February 2026), but that includes sites still under development. **Which Indian city has the most live data centre capacity?** Mumbai, including Navi Mumbai, with 890 MW live in H1 2026, about half of India's total (Cushman & Wakefield). Chennai (214 MW), Delhi NCR (183 MW), Bengaluru (162 MW), Hyderabad (151 MW) and Pune (133 MW) follow, with Kolkata at 22 MW. Mumbai also has the most capacity under construction (281 MW) and planned to 2030 (1,445 MW). **How much spare data centre space is there in India?** Colocation vacancy across the top 7 cities was 14.9% in H1 2026, down from 18.7% in H1 2025 (Cushman & Wakefield). Mumbai was the tightest at 7.9%, followed by Delhi NCR at 10.9%, while Chennai (26.7%), Bengaluru (28.0%) and Hyderabad (28.0%) had more room. For GPUs, also check when a hall is energised at full load and when liquid cooling is commissioned. **Why do Indian data centre operators' MW figures not match?** Operators measure differently. Some quote IT load (the power available to servers), some quote total power, which can be 30 to 50% higher, and some count sites still under construction. For example, CtrlS cites more than 370 MW without stating the basis (company figure, checked Oct 2026), while ICRA puts its operational capacity at 150 MW (October 2026). Ask every operator whether a figure is IT load or total power, and whether it is live. **Which operators have the biggest expansion plans in India?** Several have large pipelines, but these are plans, not live capacity. ICRA expects CtrlS to reach 950 to 1,000 MW by FY2029 (October 2026), AdaniConneX reports 960+ MW contracted against 65.4 MW operational (July 2026), and Nxtra by Airtel targets 1 GW (July 2026). Princeton Digital Group reports about 1 GW operating plus planned and US$2.5bn committed to India (March 2026). The gigawatt-scale announcements are listed in the projects tracker on the guide's main page. **What should we ask an Indian data centre operator before committing GPU space?** Ask for two dates in writing: when the hall is energised (power actually switched on) at full sanctioned load, and when liquid cooling is commissioned. Do not plan around the press-release date, because several Indian campuses have slipped by a year or more. Also confirm whether the quoted MW is IT load or total power. ### Is India a good data centre market, and what are the risks? URL: https://flexicloud.co/india-data-centers/why-india Last modified: 2026-10-23 India produces nearly 20% of the world's data but hosts about 3% of its data centres (Jun 2025). Forecasts to 2030, demand data and the risks to plan for. India produces nearly 20% of the world's data but hosts about 3% of its data centres (Jun 2025). Most 2026 forecasts see India reaching 4 to 6 GW of capacity by 2029-2030. The risks include slow grid connections and unbuilt plans: 76% of future commitments are early-stage (Cushman & Wakefield, 2026). Key facts: - India produces nearly 20% of the world's data but hosts about 3% of its data centres by count. (as of Jun 2025; Economic Survey 2025-26 (citing Nasscom, World Bank)) - About 1.6 to 1.8 GW of data centre capacity is live in India, roughly 1.5 to 1.7% of global capacity by MW (FlexiCloud estimate against about 103 GW worldwide). (as of 2025-26 (India mid-2026; global base 2025); JLL; Cushman & Wakefield) - Cushman & Wakefield and Invest India forecast 5.65 GW by end-2030 if every planned project is delivered on time; JLL forecasts 6 GW by 2029. (as of Oct 2026 (Cushman & Wakefield); Aug 2026 (JLL); Cushman & Wakefield and Invest India; JLL) - Mobile data traffic reached 285,376 PB in FY2025-26, up 24.7% in a year, with 28.4 GB a month per data subscriber in Apr to Jun 2026. (as of FY2025-26; Apr to Jun 2026; TRAI) - 76% of indicative future data centre commitments are early-stage or land-banked, so announced capacity is not the same as supply. (as of 2026; Cushman & Wakefield) **Why is India seen as a strong data centre market this decade?** India produces nearly 20% of the world's data but hosts about 3% of its data centres by count (Economic Survey 2025-26, citing Nasscom and the World Bank, Jun 2025). Demand keeps rising: India had 1.11 billion internet subscriptions on 30 Jun 2026, including about 99 million machine connections, and mobile data traffic grew 24.7% in FY2025-26 (TRAI). Building is also cheaper than in many markets, at about US$7.4m per MW against US$9.6m in Malaysia and US$15.2m in the USA (Cushman & Wakefield, 2026). **How much data centre capacity could India have by 2030?** Most forecasts land between 4 and 6 GW by 2029-2030, up from about 1.6 to 1.8 GW live in mid-2026. CareEdge Ratings sees about 4 GW by FY2030 (Mar 2026), JLL 6 GW by 2029 (Aug 2026), and Cushman & Wakefield with Invest India 5.65 GW by end-2030 if every planned project is delivered on time (Oct 2026). AI-heavy scenarios reach about 8 GW; Wood Mackenzie's 12 GW (Jul 2026) is an outlier that uses a different definition. The spread comes mostly from whether early-stage and land-banked projects are counted. **What are the biggest risks for a data centre or GPU cloud project in India?** Getting grid power connected on time, water, unsettled rules, and announcements that are not yet built: Cushman & Wakefield says 76% of indicative future commitments are early-stage or land-banked (2026). Older halls may not suit AI density, and the rupee moved from about ₹85.8 (Jun 2025) to ₹95.3 (Oct 2026) per US dollar (ECB). The Economic Survey 2025-26 also models finance as the first bottleneck and GPU supply as the main one after that. **Does India have enough power for new data centres?** At national level, largely yes: only 0.03% of energy demand went unmet in FY2025-26 (CEA), and 55% of installed capacity was non-fossil on 30 Sep 2026 (CEA), though coal still produces most of the electricity. The problem is at the site: national supply is not the same as substation capacity at a given campus (Cushman & Wakefield, Oct 2026), and the Economic Survey 2025-26 models a 12-quarter lag to add grid capacity. Lease only halls with sanctioned and energised power, or get contractual energisation dates with penalties. **Are India's data centre rules and incentives settled?** Not fully, as of October 2026. Most duties under the DPDP Rules (India's privacy law, notified Nov 2025) start on 13 May 2027, and the amended 2047 data centre tax exemption became law on 17 Aug 2026 (Act No. 21 of 2026), but its prescribed conditions are not yet published. Cushman & Wakefield still cited a 2020 draft national data centre policy in Oct 2026. This is general information, not legal or tax advice; take Indian legal and tax advice before relying on any incentive. ### How much are AWS, Microsoft and Google investing in India, and why? URL: https://flexicloud.co/india-data-centers/hyperscaler-investment Last modified: 2026-10-24 Amazon, Microsoft and Google have pledged more than US$56bn for AI and cloud in India, mostly for 2026 to 2030. What is live as of Oct 2026, and why. Amazon, Microsoft and Google have pledged more than US$56bn for AI and cloud in India (Oct 2026). Most of it is for 2026 to 2030, though Microsoft's Hyderabad region went live on 6 Aug 2026. The reason: India produces nearly 20% of the world's data but hosts about 3% of its data centres (Jun 2025). Key facts: - Amazon plans US$48bn across all its India businesses for 2026 to 2030, of which US$21bn+ is for AI and cloud. (as of 25 Jun 2026; Amazon India) - Microsoft has pledged US$20.5bn for 2025 to 2029, and its new three-zone Hyderabad region went live on 6 August 2026. (as of 6 Aug 2026; Microsoft) - Google plans about US$15bn for 2026 to 2030 on a gigawatt-scale AI hub in Visakhapatnam with AdaniConneX and Airtel. Ground was broken on 28 April 2026. (as of 28 Apr 2026; Google India) - Adani Group has pledged US$100bn for AI infrastructure to 2035. This includes capacity built for Google and Microsoft, so do not add it to their pledges. (as of 17 Feb 2026; Adani Enterprises) - A US AI cloud company will be the sole tenant of 240 MW at AdaniConneX Taloja, Navi Mumbai, with an option for 240 MW more. The first phase is due mid-2028. (as of 6 Oct 2026; Inc42) **How much have AWS, Microsoft and Google pledged to invest in India?** More than US$56 billion for AI and cloud, mostly for 2026 to 2030. Amazon plans US$48bn across all its India businesses from 2026 to 2030, of which US$21bn+ is for AI and cloud (25 Jun 2026). Microsoft has pledged US$20.5bn for 2025 to 2029 (9 Dec 2025), and Google about US$15bn for 2026 to 2030 (14 Oct 2025). **Which of the new hyperscaler projects in India are live or being built?** As of October 2026, Microsoft's new three-zone Hyderabad region is live; it opened on 6 August 2026. Google broke ground on its gigawatt-scale Visakhapatnam campus, built with AdaniConneX and Airtel, on 28 April 2026. AWS broke ground on a new Hyderabad data centre at Bharat Future City on 15 July 2026. Most of the pledged money is still a plan. **Why are hyperscalers investing so much in India?** Three reasons stand out. Demand inside India: data rules keep regulated workloads in the country, and India is in Google's top 3 countries for Gemini (Feb 2026). Competition: Google (14 Oct 2025), Microsoft (9 Dec 2025) and Amazon (10 Dec 2025) announced pledges within two months of each other. Cost and policy: a tax exemption for export cloud, cheaper builds than most markets, talent and clean power. The wider market case, including India's share of the world's data, is on the Why India page of this guide. **Who else is investing in AI data centres in India?** Indian groups have pledged more: Adani US$100bn to 2035 (17 Feb 2026), and Reliance about ₹10 lakh crore over 7 years (reported by TechCrunch on 19 Feb 2026; not found in a Reliance filing). AirTrunk plans US$30bn by 2030 (5 Jun 2026). Reliance will build a 168 MW built-to-suit AI data centre for Meta at Jamnagar, due within two years (RIL, 10 Jun 2026), a US AI cloud company took 240 MW at Taloja with AdaniConneX (6 Oct 2026), and TCS HyperVault has OpenAI as its first customer, starting at 100 MW (Feb 2026). **Can these investment pledges be added up to one total?** No, there is no reliable single total. The pledges cover different periods, Amazon's figure includes businesses beyond data centres, and Adani's includes capacity built for Google and Microsoft, so adding them double counts. The cleanest figure is the big three's AI and cloud pledges: more than US$56 billion, mostly for 2026 to 2030, as of October 2026. **What does hyperscaler investment mean for a GPU cloud entering India?** Watch where hyperscalers and their anchor tenants build, because power, fibre and talent follow them. They are taking power and land at the biggest campuses first, and about 1.15 GW is pre-leased, roughly double the earlier level (Cushman & Wakefield, APAC Investment Landscape 2026). Expect them to compete with you for the same liquid-cooled halls, so reserve early and in writing. Sovereignty rules and in-country inference also create local GPU demand that offshore capacity cannot serve. ### How do data centre power, build and colocation costs in India compare with other countries? URL: https://flexicloud.co/india-data-centers/power-costs Last modified: 2026-10-23 Data centre costs in India vs other countries: FY2026-27 power tariffs by state, build cost per MW and colocation rent per kW, all sourced and dated. Grid power for large users in India costs about ₹6.2 to ₹10.9 per kWh, a mid-range price (FY2026-27 tariffs). Building is cheap: about US$7.4m per MW, against US$15.2m in the USA (Cushman & Wakefield, 2026). Mumbai colocation costs US$73 to US$105 per kW a month (H1 2025), against US$330 to US$475 in Singapore (Q1 2026). Key facts: - In Maharashtra (MSEDCL), large industrial users on high-tension supply pay an average ₹10.88 per kWh (US$0.114). If billed as commercial, they pay ₹16.74 (US$0.176), about 54% more. (as of FY2026-27; MERC (tariff order, 25 Mar 2026)) - Singapore's regulated tariff for large users works out at S$0.235 to S$0.246 per kWh all-in (US$0.184 to US$0.192). (as of Q4 2026; SP Group) - Building a data centre in India costs about US$7.4m per MW (about ₹70 crore; range US$5.9m to US$9.0m), excluding land and IT equipment. Malaysia is US$9.6m and the USA US$15.2m. (as of 2026; Cushman & Wakefield cost guides) - Mumbai colocation rent is ₹7,000 to ₹10,000 per kW a month (US$73 to US$105) for 250 kW to 1 MW, and ₹6,000 to ₹7,500 (US$63 to US$79) for 5 MW or more. (as of H1 2025; Cushman & Wakefield) - Colocation asking rents elsewhere: Northern Virginia US$190 to US$235, Frankfurt US$235 to US$265 and Singapore US$330 to US$475 per kW a month. (as of Q1 2026; CBRE) **How much does grid power cost a data centre in India?** On FY2026-27 industrial tariffs, indicative all-in prices run from about ₹6.2 per kWh (US$0.065) in Andhra Pradesh to ₹10.9 (US$0.114), Maharashtra's average billing rate for high-tension industry. Uttar Pradesh is about ₹6.6 to ₹7.7, Karnataka about ₹7.5 and Telangana about ₹7.8 to ₹8.8, while Tamil Nadu is about ₹9.0 on FY2025-26 rates. Our state estimates assume a flat 24x7 load at 85% load factor and exclude electricity duty, fuel adjustment and open access charges. **Is data centre power in India cheaper than in Singapore, Europe or the US?** It is cheaper than Singapore and Europe, and in the same range as US states. Singapore's regulated rate is US$0.184 to US$0.192 per kWh (SP Group, Q4 2026), and users taking 70 to 150 GWh a year pay an average US$0.201 in Ireland, US$0.154 in Germany and US$0.140 in the Netherlands (Eurostat, H1 2026). US industry pays an average US$0.101 in Virginia and US$0.071 in Texas (US EIA, Jul 2026). Foreign figures use different bases, so treat them as a guide, not a like-for-like quote. **Why does the tariff category matter so much for a data centre in India?** Each state distribution company (DISCOM) decides whether your connection is billed as industrial or commercial, and the gap can be large. In Maharashtra, high-tension (HT) commercial supply averages ₹16.74 per kWh against ₹10.88 for HT industry in FY2026-27, about 54% more (MERC). In Karnataka the current order makes HT commercial slightly cheaper than industry, and several state data centre policies, such as Tamil Nadu's 2021 policy, help secure industrial treatment. This is general information, not legal advice: get the category confirmed in writing for each connection. **How much does it cost to build a data centre in India per MW?** About US$7.4m per MW (about ₹70 crore), within a range of US$5.9m to US$9.0m, excluding land and IT equipment (Cushman & Wakefield, 2026). On the same basis Malaysia costs US$9.6m, the USA US$15.2m and Japan US$16.0m, so a 100 MW build in India costs about US$220m less than in Malaysia. Turner & Townsend puts Mumbai at US$6.64m (cited Feb 2026) against US$14.5m in Singapore (Nov 2025); compare within one source, as specifications differ. Indian tender prices are forecast to rise 3 to 5%, driven mainly by labour and specialist mechanical, electrical and plumbing (MEP) work (Turner & Townsend, 13 Aug 2026). **What does colocation cost in India compared with Singapore or Northern Virginia?** Mumbai rents were ₹7,000 to ₹10,000 per kW a month (US$73 to US$105) for 250 kW to 1 MW and ₹6,000 to ₹7,500 (US$63 to US$79) for 5 MW or more, while Chennai, Hyderabad and Delhi NCR ranged from ₹6,000 to ₹8,500 (Cushman & Wakefield, H1 2025). CBRE's Q1 2026 asking rents were US$190 to US$235 in Northern Virginia, US$235 to US$265 in Frankfurt and US$330 to US$475 in Singapore. The Indian figures are about a year older, and whether power is included differs by market. Space for AI training commands about a 60% lease premium (JLL, Jan 2026). **Can a data centre in India buy green power directly, and what does open access add?** Generally yes. Cushman & Wakefield and Invest India (2026) report green energy open access from 100 kW, with deemed approval after 15 days; we have not checked this against each state's rules. Buying from a third party adds a cross-subsidy surcharge of ₹2.11 per kWh in Maharashtra, ₹1.99 in Tamil Nadu and ₹1.37 to ₹1.71 in Andhra Pradesh (Vizag DISCOM), plus wheeling of ₹0.81 to ₹1.04 (FY2025-26 to FY2026-27), though captive or group captive set-ups (26%+ ownership and 51%+ consumption) are exempt from cross-subsidy and additional surcharges. This is general information, not legal advice, and we found no sourced all-in cost for round-the-clock green power. ### Which subsea cables land in India, and what is coming? URL: https://flexicloud.co/india-data-centers/subsea-cables Last modified: 2026-10-23 About 15 international subsea cables are live in India, mostly in Mumbai and Chennai, with 5 more planned. Every live and coming cable by city (Oct 2026). About 15 international subsea cables are live in India, mostly landing in Mumbai and Chennai (TeleGeography, 14 Oct 2026). Five more are planned, starting with SEA-ME-WE 6, India Europe Xpress and Raman in 2027. New east-coast landings at Visakhapatnam and Machilipatnam are mostly 2027 to 2030 projects, and the IAX branches listed there are not confirmed to carry traffic. Key facts: - About 15 international subsea cables are live in India, plus 2 domestic cables to the islands. 5 more systems are planned. (as of 14 Oct 2026; TeleGeography) - Mumbai has about 9 to 11 international landing stations and about 11 to 12 live international cables. Chennai has about 6 stations and 6 live cables. (as of 14 Oct 2026; FlexiCloud tally from TeleGeography, operator releases and TRAI) - The latest official figure: India's lit international subsea capacity was 138,606 Gbps, with 111,111 Gbps (80%) activated. (as of end-2022 (published 19 Jun 2023); TRAI) - SEA-ME-WE 6 landed in Mumbai (Dec 2024) and Chennai (Feb 2025), with 126 to 132 Tbps of design capacity. Service is expected in 2027. (as of Oct 2026; Bharti Airtel; TeleGeography) - Cable repairs in Indian waters have taken 3 to 5 months, because repair ships are based abroad. (as of Jun 2023; TRAI) **How many subsea cables land in India?** About 15 international subsea cables are live in India, plus 2 domestic cables to the islands: CANI to Andaman and Nicobar (2020) and KLI to Lakshadweep (2024), as of 14 Oct 2026 (TeleGeography). TRAI's last official count (Jun 2023) was 17 cables, and no single official list exists. TRAI also reported 138,606 Gbps of lit international capacity at end-2022, with 80% of it activated. **Where are India's cable landing stations?** Mostly in Mumbai and Chennai. Mumbai has about 9 to 11 international landing stations and about 11 to 12 live international cables; Chennai has about 6 stations and 6 live cables (our tally, 14 Oct 2026). Kochi, Thiruvananthapuram and Tuticorin have one international cable each. **Which new subsea cables are coming to India, and when?** SEA-ME-WE 6, India Europe Xpress (IEX) and Raman are expected in 2027, all landing in Mumbai, with SEA-ME-WE 6 also in Chennai (TeleGeography, Oct 2026). Raman's Red Sea section was delayed as of Jan 2026. 2Africa Pearls landed in Mumbai in Mar 2025 and Meta expects it to go live in 2026. I-2SEA is targeted for Q4 2029, and Meta's Project Waterworth has no date yet. **Will India's east coast get new cable landings beyond Chennai?** Yes, at Visakhapatnam and Machilipatnam in Andhra Pradesh, but not soon. Google has announced a Visakhapatnam gateway with Airtel building the landing station, with no cable names or dates yet (Feb 2026). I-2SEA contracts were signed on 2 Jul 2026 for landings at Machilipatnam and South Chennai, with service targeted for Q4 2029. Treat both as 2027 to 2030 options. TeleGeography also lists IAX branches at Machilipatnam and Digha, but we have not confirmed they carry commercial traffic. **How many cable routes should a GPU cloud in India buy?** At least two, on different cables. Most of Mumbai's westbound routes cross the Red Sea, so make sure one route avoids it. Cable repairs in Indian waters have taken 3 to 5 months (TRAI, 2023), so a single cable is a single point of failure. Mumbai has the most cables and landing operators, and Chennai is the direct route to Singapore. Which city is best connected overall is covered on the connectivity page of this guide. **What are the rules for landing a subsea cable in India?** TRAI proposed easier landing rules, cable corridors and incentives for new landing cities in June 2023. In February 2025 they were folded into a new long-distance authorisation, and new authorisation rules dated 23 Jun 2026 (as reported by Airtel) merge national and international long distance. We have not confirmed whether the landing incentives made it into those rules. This is general information, not legal advice; confirm the current position with Indian telecom counsel before you plan around it. ### How are India's data centres connected? URL: https://flexicloud.co/india-data-centers/connectivity Last modified: 2026-10-23 Internet exchanges, cloud on-ramps, fibre carriers and latency between Indian data centres, with network counts and port prices as of Oct 2026. Mumbai is India's connectivity hub: 451 networks were connected to DE-CIX there on 14 Oct 2026. AWS, Microsoft Azure, Google Cloud and Oracle all offer private cloud on-ramps in India (Oct 2026). A public internet round trip from Mumbai to Delhi took a median 27 milliseconds (29 Sep to 13 Oct 2026). Key facts: - DE-CIX had 451 networks connected in Mumbai, against 105 in Delhi and 78 in Chennai. (as of 14 Oct 2026; PeeringDB) - Equinix MB1 in Mumbai is India's most interconnected building, with 261 networks present. (as of 14 Oct 2026; PeeringDB) - A 10G internet exchange port costs ₹50,000 (US$525) a month at Extreme IX and ₹49,900 (US$524) at NIXI metro sites, before tax. (as of 14 Oct 2026 (NIXI tariff in force from 1 Jul 2025); Extreme IX and NIXI) - An AWS Direct Connect 10G port costs US$2.25 an hour, about US$1,642 (₹1.56 lakh) a month. (as of 14 Oct 2026; Amazon Web Services) - Median public internet round trip: 23 milliseconds Mumbai to Chennai, 27 milliseconds Mumbai to Delhi. (as of 29 Sep to 13 Oct 2026; WonderNetwork) **Which Indian city is best connected for a data centre or GPU cloud?** Mumbai, in most cases. It has the busiest internet exchanges and the most connected buildings: on 14 Oct 2026, DE-CIX had 451 networks in Mumbai and Equinix MB1 alone hosted 261 (PeeringDB). Delhi NCR is second by networks at the exchanges (DE-CIX 105, Extreme IX 175, NIXI 114), while Chennai is the natural second site for subsea routes to Singapore. We suggest a carrier-neutral Mumbai building with access to at least two exchanges and two cloud on-ramps. **How much does an internet exchange port cost in India?** At Extreme IX, a 10G port costs ₹50,000 (US$525) a month and a 100G port ₹2,50,000 (US$2,625), before tax (14 Oct 2026). NIXI's metro rates are ₹49,900 (US$524) and ₹2,47,333 (US$2,597) a month, with cheaper annual billing (10G ₹5,80,000 a year) and edge nodes at about 2.5 times the price (tariff from 1 Jul 2025). DE-CIX India does not publish its prices. **Can I connect privately to AWS, Azure, Google Cloud or Oracle in India?** Yes. All four offer private on-ramps in India: AWS Direct Connect, Azure ExpressRoute, Google Cloud Interconnect and Oracle FastConnect (provider pages read 8 to 14 Oct 2026). Mumbai has on-ramps for all four, for example at Equinix MB1 (AWS, Google), Sify Rabale (AWS, Oracle) and Tata Communications (Azure, Google). Put on-ramps in two different buildings so one failure does not cut you off; at about US$1,642 (₹1.56 lakh) a month per AWS 10G port (US$2.25 an hour), budget for both. **What is the latency between Mumbai, Delhi and Chennai?** On the public internet, the median round trip was 27 milliseconds from Mumbai to Delhi and 23 from Mumbai to Chennai (WonderNetwork, 29 Sep to 13 Oct 2026). Microsoft's own backbone measured 20 milliseconds from Pune to Chennai over the 30 days to 30 Jul 2026. The physical minimum in fibre is about 11 and 10 milliseconds. Some public routes detour badly, which is why private links and direct peering matter. **Do I need a telecom licence to run a GPU cloud in India?** In our reading, not if you only sell compute: colocating, buying cross-connects and peering with your own ASN (the network number you need to exchange traffic directly) need no telecom licence. Selling internet access, IP transit or inter-city circuits needs a Department of Telecommunications authorisation, under new rules dated 23 Jun 2026 as reported by Airtel. Since 15 Sep 2021, a telecom licensee can be 100% foreign-owned under the automatic route, with safeguards. This is general information, not legal advice; confirm with Indian telecom counsel before you contract. **Should a GPU cloud lease dark fibre between its Indian sites?** Usually not directly. Dark fibre, ducts and rights of way are leased by registered infrastructure providers (IP-I), and only to licensed operators, so a GPU cloud normally buys wavelengths or managed fibre from licensed carriers, ideally two carriers on different routes. Large carriers include Bharti Airtel, with 542,570 route km of fibre in India (Jun 2026), and PowerGrid, with 100,000+ km (Oct 2026). This is general information, not legal advice. ### How can a GPU cloud (neocloud) enter India? URL: https://flexicloud.co/india-data-centers/neocloud-guide Last modified: 2026-10-23 How a GPU cloud (neocloud) can enter India: entry models, CERT-In, DPDP and US export rules, IndiaAI prices, state incentives and a 14-point checklist. Colocation is fastest: with liquid-cooled space ready, first GPUs can go live in weeks to a few months (our estimate). In our reading, CERT-In rules apply from day one, including 6-hour reporting of cyber incidents (since 2022). The 2047 tax exemption probably does not fit a GPU cloud that owns its GPUs; this is general information, not legal or tax advice. Key facts: - In our reading, CERT-In's directions, which name data centres and cloud providers, require cyber incidents to be reported within 6 hours and logs to be kept for 180 days; CERT-In's FAQ allows the logs to be stored outside India if you can produce them in reasonable time. (as of Oct 2026 (in force since 28 Apr 2022; FAQ May 2022); CERT-In) - The DPDP Rules were notified in Nov 2025. Consent-manager provisions start on 13 Nov 2026 and most other duties on 13 May 2027. (as of Nov 2025; PIB / MeitY) - India is in US Country Group A:5. A US licence is still needed for customers headquartered in, or owned from, China, Macau or another D:5 country, and the US suspension of its affiliates rule ends on 9 Nov 2026 unless extended. (as of Oct 2026 (BIS guidance of 31 May 2026); US BIS; Federal Register) - In our reading, the 2047 data centre tax exemption (Income-tax Act 2025, Schedule IV, item 13C) applies only if the foreign company does not own or operate any of the data centre's physical infrastructure or resources and sells to users in India through an Indian reseller company. The amendments became law on 17 Aug 2026 (Act No. 21 of 2026); the prescribed conditions are not yet published. (as of Oct 2026 (amended Act in force from 1 Apr 2026); Gazette of India; PRS Legislative Research) - IndiaAI's reference on-demand rate for an 8-GPU H100 SXM instance is ₹153 (US$1.61) per GPU-hour. It is open only to eligible users (startups, researchers, academia and government), who can get up to 40% off. (as of 13 Mar 2026; IndiaAI / MeitY (Ready Reckoner)) **What is the fastest way for a GPU cloud to start operating in India?** Colocation: lease racks, a cage or a suite in an existing Indian data centre. If liquid-cooled, high-density space is available, first GPUs can be live in weeks to a few months, against about 6 to 12 months after handover for a powered shell and about 18 to 30 months for build-to-suit (FlexiCloud estimates, Oct 2026). The catch is that space for 100 kW+ racks is scarce, so get the grid energisation date in writing. **Which Indian rules apply to a foreign GPU cloud?** In our reading, CERT-In's 2022 directions apply from day one: report incidents within 6 hours, keep logs for 180 days (CERT-In's FAQ allows storage outside India if you can produce them in reasonable time) and keep verified customer records for 5 years. The DPDP Rules, notified in Nov 2025, bring consent-manager provisions from 13 Nov 2026 and most other duties from 13 May 2027. Serving payments firms means keeping their payment data in India (RBI, since 2018); SEBI-regulated firms need their data kept in India (since 2023), and SEBI work also needs MeitY empanelment. This is general information, not legal advice; confirm each point with Indian counsel. **Do we need a US export licence to serve Indian customers with NVIDIA GPUs?** Generally no, as of October 2026: India is in US Country Group A:5, so most Indian customers need no US licence for advanced NVIDIA GPUs. A licence is needed for any customer headquartered in, or ultimately owned from, China, Macau or another D:5 country (BIS guidance, 31 May 2026), and end-use, end-user and Entity List checks always apply. The US suspension of its affiliates rule ends on 9 Nov 2026 unless extended, which could tighten ownership screening. This is general information, not legal advice; confirm with US export-control counsel before each order. **Does the 2047 data centre tax exemption help a GPU cloud that owns its GPUs in India?** Probably not. The exemption (Income-tax Act 2025, Schedule IV, item 13C, added by the Finance Act 2026) requires the foreign company not to own or operate any of the data centre's physical infrastructure or resources, to sell to users in India only through an Indian reseller company, and to buy its data centre services from a data centre run by an Indian company, which may own or lease it. The amendments became law on 17 Aug 2026 (Act No. 21 of 2026), but as of Oct 2026 the prescribed conditions had not been published. This is general information, not tax or legal advice; confirm with an Indian chartered accountant or tax counsel. **How do IndiaAI GPU prices affect what we can charge in India?** IndiaAI's reference on-demand rates (13 Mar 2026) on 8-GPU SXM instances are ₹153 (US$1.61) per GPU-hour for H100, ₹140.63 (US$1.48) for H200 and ₹323 (US$3.39) for B200. Only eligible users (startups, researchers, academia and government) can use them, and they can get up to 40% off. As of 2 Sep 2026, 14 empanelled bidders offered 45,081 accelerators (capacity offered, not necessarily live), and these rates set price expectations for startups and academia. Decide early whether to bid, partner or stay out. **Which Indian states offer data centre incentives?** Eleven states had data centre policies in force as of Sep 2026 (Cushman & Wakefield / Invest India). We verified two: Tamil Nadu's 2021 policy offers an industrial power tariff and a 5-year electricity tax waiver for ₹500 crore (US$52.5m) or more of fixed assets, with at least 30% renewable energy for the power incentives, and Uttar Pradesh's 2026 policy (notified 9 Jul 2026) offers units of 2 to 40 MW a 7% capital subsidy of up to ₹20 crore. Most incentives need approval before you invest, so apply first and check the current policy in other states. This is general information, not legal advice. ## Contact - Contact page, with the enquiry form: https://flexicloud.co/contact - Book a call: https://flexicloud.co/contact#book-a-call - Office address, phone and email: https://flexicloud.co/about#contact