Skip to content

Vera Rubin hardware: specs, NVLink, memory and the cable-free build

22 November 2026Vinod Chacko
Vera Rubin hardware: specs, NVLink, memory and the cable-free build

This is the hardware deep dive in our three-part guide to NVIDIA Vera Rubin. For the short answers on price, timing and whether to wait, start with Vera Rubin explained. For costs, benchmarks and supply, see Vera Rubin price, performance and timing.

Figures are as of 24 October 2026. NVIDIA has revised several Rubin specs since launch, so we use its current product pages and say where earlier figures differ.

Vera Rubin vs Blackwell: the hardware in numbers

Four things change for buyers. Memory bandwidth per GPU rises about 2.4 times in the NVL72 rack, and about 2 times on NVIDIA's figure for 8-GPU servers. Dense low-precision compute (the FP4 and FP8 number formats used to run models) rises about 2.3 times in FP4 and 3.5 times in FP8, against a GB300 GPU.

NVLink rises about 1.7 times, and so does power per 8-GPU server. Memory capacity stays at 288 GB per GPU, the same as the B300.

An 8-GPU server: HGX B300 against HGX Rubin NVL8

This is the format we source for customers. Figures are NVIDIA's published peak specs for each 8-GPU system. FP4, FP8 and BF16 are number formats for AI maths: the fewer the bits, the faster the GPU runs.

FP4 and FP8 are compact formats that run models fastest; BF16 is a common format for training. FP64 is the high-precision format used in science. "Dense" figures count no shortcuts. NVIDIA's headline figures are often "sparse": they count a shortcut that only helps when a model has been pruned in a particular pattern.


HGX B300

HGX Rubin NVL8

Change

GPU memory

288 GB of HBM3e per GPU

288 GB of HBM4 per GPU

Same size

Memory bandwidth

8 TB/s per GPU

130 TB/s for 8 GPUs, about 16 TB/s each

About 2x

FP4 compute, dense

108 PF

270 PF

About 2.5x

FP8 compute, dense

36 PF

130 PF

About 3.6x

BF16 compute, dense

18 PF

32 PF

About 1.8x

FP64 (double precision)

10 TF

265 TF native

Restored, close to the B200's 296 TF. NVIDIA also lists 200 TF per GPU for emulated FP64 matrix maths

NVLink switch bandwidth

14.4 TB/s

24 TB/s

About 1.7x

Network per GPU to other servers

800 Gb/s (DGX B300)

800 Gb/s (DGX Rubin NVL8)

Same

Host CPUs

x86, chosen by the server maker (DGX B300: two Intel Xeon)

x86, chosen by the server maker (DGX Rubin NVL8: two Intel Xeon), or one Vera

Mostly the same

Power per server

About 14 kW (DGX B300)

About 24 kW (DGX Rubin NVL8)

About 1.7x

Source: NVIDIA's HGX and DGX pages, read 24 October 2026. DGX is NVIDIA's own-brand version of the 8-GPU server. PF is petaflops and TF teraflops. TB/s is terabytes per second; network speeds are in Gb/s or Tb/s (gigabits or terabits per second), and 1 Tb/s is one eighth of 1 TB/s. Ratios are our arithmetic.

HGX Rubin NVL8 against HGX B300

Memory size per GPU stays at 288 GB. Power is NVIDIA’s DGX figure: about 24 kW against about 14 kW.

Source: NVIDIA’s HGX and DGX pages, read 24 October 2026; ratios are our arithmetic

Two things stand out. First, NVIDIA's own table lists 2 TB of HBM4 for 8 Rubin GPUs, slightly less than the 2.1 TB it lists for the HGX B300.

Server makers list 2.3 TB (8 times 288 GB). If your model does not fit on a B300 node today, Rubin will not change that; see our GPU memory guide for sizing.

Second, power rises by about 70%, and every 8-GPU Rubin server design we found cools its GPUs with liquid. That narrows the data centre halls that can take them.

A rack: GB300 NVL72 against Vera Rubin NVL72


GB300 NVL72

Vera Rubin NVL72

Change

GPUs and CPUs

72 Blackwell Ultra GPUs, 36 Grace CPUs

72 Rubin GPUs, 36 Vera CPUs

Same count

GPU memory (HBM)

20 TB

20.7 TB

About the same

GPU memory bandwidth

576 TB/s

1,400 TB/s

About 2.4x

CPU memory (LPDDR5X)

Up to 17 TB

Up to 54 TB

About 3x on paper

FP4 compute, dense / sparse

1,080 / 1,440 PF

2,520 / 3,600 PF

About 2.3x / 2.5x

FP8 compute, dense

360 PF

1,260 PF

About 3.5x

NVLink per GPU

1.8 TB/s

3 TB/s

About 1.7x

NVLink per rack

130 TB/s

216 TB/s

About 1.7x

Network per GPU to other racks

800 Gb/s

1.6 Tb/s

2x

Rack power

About 135 to 150 kW (estimates)

156 to 227 kW, by power profile (NVIDIA)

About 1.5x to 1.7x at full power

Cooling

About 80% liquid

Compute trays 100% liquid

No fans in the trays

Source: NVIDIA's Vera Rubin NVL72 and GB300 NVL72 pages and the Vera Rubin NVL72 datasheet (October 2026). GB300 rack power is from press and analyst estimates, as NVIDIA does not list it; the cooling share is Huang's figure. Ratios are our arithmetic.

NVIDIA's headline 50 petaflops per Rubin GPU is the sparse figure; the dense figure is 35. We compare dense with dense throughout.

The Vera CPU

Vera has 88 NVIDIA-designed Olympus cores and 176 threads, against 72 cores for Grace. NVIDIA lists up to 1.5 TB of memory per CPU at 1.2 TB/s (Grace: 480 GB at 512 GB/s). Its link to the GPUs runs at 1.8 TB/s, twice Grace's 900 GB/s.

That memory matters for AI serving, because the GPUs can park the working memory of long conversations (the KV cache) in it. One caution: TrendForce and SemiAnalysis report that early systems ship with about half that CPU memory, because of the memory shortage.

Specs NVIDIA has revised

Read the current product page, not the launch slides. NVIDIA has changed two headline figures without explaining why:

  • Memory bandwidth per GPU: 13 TB/s when Rubin was first shown in March 2025, up to 22 TB/s at CES in January 2026, and 19.2 TB/s on NVIDIA's Vera Rubin NVL72 page and October 2026 datasheet today.
  • Memory bandwidth on other NVIDIA pages: NVIDIA's HGX page still shows 22 TB/s in its per-GPU table, while its 8-GPU system figure, 130 TB/s, works out to about 16 TB/s per GPU, about twice Blackwell's 8 TB/s (our arithmetic).
  • The DGX Rubin NVL8 page: Even the DGX Rubin NVL8 page disagrees with itself: its text says 176 TB/s for 8 GPUs (22 each), its spec table 130 TB/s.
  • NVLink 6 per GPU: 3.6 TB/s at CES (260 TB/s per rack), and 3 TB/s (216 TB/s per rack) on NVIDIA's current US NVLink and Vera Rubin NVL72 pages and its October 2026 datasheet. Only the NVLink page marks the figure as preliminary.
  • NVLink 6 on other NVIDIA pages: NVIDIA's HGX page (in its per-GPU table) and some of its regional NVLink pages, including India's, still show 3.6 TB/s.

Rubin memory bandwidth per GPU: the figure has moved

The HGX per-GPU figure is 130 TB/s for 8 GPUs, divided by 8 (our arithmetic).

Source: NVIDIA product pages and October 2026 datasheet

So before you sign for Rubin capacity, ask the server maker or provider for the exact figures of the system you are getting, in writing.

One product is off the 2026 plan: Rubin CPX, a lower-cost GPU meant for reading very long inputs, was pulled at GTC in March.

Need GPUs in the next 6 to 12 months? See B300 and B200 server prices, or tell us your workload and we will confirm the lead time for your region.

How the Vera Rubin rack and tray are built

The biggest change is inside the NVL72 compute tray, the flat drawer that holds the chips. Its fans and almost all its internal cables are gone, and its liquid tubes drop from six to two.

NVIDIA says robots now build one in about five minutes, against about two hours by hand for Grace Blackwell. The rack around it changes much less.

Inside the tray: almost no cables, fewer hoses, no fans

Each Vera Rubin NVL72 compute tray holds two Vera CPUs and four Rubin GPUs. On a Grace Blackwell tray, dozens of internal cables (43 by Huang's count; Fast Company, which saw both trays side by side, counted 50 or 60) and six liquid tubes had to be fitted by hand.

On Vera Rubin, a circuit board across the middle of the tray (a midplane) and board-to-board connectors replace the internal signal cables. Quick-release coolant connectors cut the tubes from six to two. The fans are gone too, because everything on the tray is liquid-cooled, and going fanless is what freed the space for the midplane.

NVIDIA's figures for the gain:

  • Build time: by robots, not by hand. NVIDIA's own figures vary: from over 1.5 hours to 3 hours for Grace Blackwell, and from one to five minutes for Vera Rubin.
  • Right first time: Andrew Bell, NVIDIA's senior vice president of hardware engineering, told Fast Company that only about 20% of early Grace Blackwell trays were right first time. For Vera Rubin, Bell put the figure at about 95%.
  • Servicing: NVIDIA's launch release said assembly and servicing are up to 18 times faster; later NVIDIA documents say 20 times and, in its October 2026 datasheet, 90 times. The German tech site heise, citing NVIDIA, reported that swapping an NVLink switch tray takes about six minutes, against 100 minutes on Grace Blackwell.

Minutes to build a tray, or swap a switch tray

Vendor figures that nobody has measured independently. NVIDIA’s own build figures vary: over 1.5 hours to 3 hours for Grace Blackwell, one to five minutes for Vera Rubin.

Source: NVIDIA; heise, citing NVIDIA

These are vendor figures that nobody has measured independently. The direction still matters: cables were the main failure point when Grace Blackwell trays were built, and they slowed that rollout (SemiAnalysis, Tom's Hardware). A tray that is quicker to build and repair should mean faster ramps and shorter downtime when a part fails.

The rack: an evolution, not a revolution

Vera Rubin NVL72 uses the same rack footprint as the GB300 NVL72. Both follow NVIDIA MGX, the standard rack design that partners build to. So a data centre that took Grace Blackwell racks can plan for Rubin in a similar way, with more power and more coolant flow (SemiAnalysis estimates about 2 to 2.5 times Blackwell's). Each rack holds:

  • 18 compute trays and 9 NVLink switch trays;
  • a copper NVLink "spine" at the back: about 5,000 cables in four cartridges, more than 3 km of cable in total;
  • four 110 kW power shelves fed with AC power, which feed a 54 V DC busbar (a solid power rail) up the rack;
  • liquid cooling with coolant at up to 45°C.

Power is the real change: up to 227 kW per rack at full power, or 156 kW in NVIDIA's lowest-power profile, against an estimated 135 to 150 kW for a GB300 NVL72.

The rack's copper NVLink spine goes away only with Kyber, the rack NVIDIA designed for Rubin Ultra, the higher-end follow-on to Rubin. Kyber uses vertical blades, a circuit-board midplane instead of the spine, and 800 V DC power.

NVIDIA's roadmap shows it in 2027. SemiAnalysis reported in July 2026 that it has slipped into 2028; NVIDIA has not confirmed that and says its roadmap is intact.

Who builds it

More of the building moves to NVIDIA and a few large manufacturers. Server makers have traditionally bought NVIDIA's GPU boards and built their own trays.

For Vera Rubin, J.P. Morgan reported in November 2025 that NVIDIA would ship fully built and tested compute trays (what the industry calls "L10"). Partners would then assemble and test whole racks ("L11"). NVIDIA has not confirmed it.

SemiAnalysis says only three manufacturers, Foxconn, Quanta and Wistron, have automated tray lines. NVIDIA's partner list for Vera Rubin systems includes Dell, HPE, Lenovo, Supermicro, Foxconn, Inventec, QCT, Wistron, Wiwynn, Pegatron and Compal, across more than 350 factories in 30 countries.

If the L10 reports are right, Rubin racks from different brands will be more alike than before. Server makers will then compete more on delivery, integration and support than on design.

NVIDIA has also announced DSX POD, a design for prefabricated AI factories in containers that partners build and test before shipping. It says deployment at a prepared site can take as little as 14 days instead of six months, with first deliveries planned for early 2027.

8-GPU Rubin servers

The 8-GPU HGX Rubin NVL8 servers that we would source are built by server makers, not NVIDIA, and designs differ.

All five we found, from Supermicro, ASUS, Compal, Wiwynn and Gigabyte, are 2U or 3U servers (two or three standard rack units tall) with liquid-cooled GPUs. Compal lists about 24 kW per server, in line with NVIDIA's figure for its DGX Rubin NVL8.

Some are cable-free inside, such as an ASUS design with board-to-board connections; others, such as a Gigabyte server shown at GTC, still use cables for the drives. One Supermicro-built system is sold only as a complete, integrated rack, so not every design can be bought as a single node.

Is Vera Rubin really a "wireless" server?

Yes, almost certainly, but "wireless" here means cable-free inside the tray, not radio. Huang has shown the Vera Rubin compute tray at keynote after keynote since October 2025, and stressed the missing cables each time:

When

Where

What Huang said

28 October 2025

GTC Washington, DC

The tray is "completely cable-less" and 100% liquid-cooled

5 January 2026

CES, Las Vegas

Comparing trays: "43 cables, zero cables"

1 June 2026

GTC Taipei

"No cables, no hoses, no fans"

21 October 2026

GTC Berlin

"No cables, no wires"

Quotes are from third-party transcripts of the keynotes. We found no record of Huang calling the server "wireless". The word most likely comes from Huang's "no wires" line in Berlin. It also appeared in coverage of the CES talk: Taiwan's Economic Daily News used a Chinese term that translates as "going wireless".

What is still cabled or plumbed:

  • The NVLink spine: about 5,000 copper cables at the back of each rack, linking all 72 GPUs.
  • Power: AC power feeds into each power shelf, whose power supplies are still air-cooled.
  • Tray connections: NVIDIA's Andrew Bell told Fast Company the "cableless" tray still has two cables, one to power and one to the network. Each tray is also cabled to the cluster network and the management network.
  • Liquid: quick-release couplings from each tray to the rack's coolant manifolds.

The benefit is faster building and servicing, not speed: a cable-free tray does not make the GPUs compute any faster. The claim applies only to the NVL72 compute and switch trays. NVIDIA's HGX page makes no cable-free claim for 8-GPU servers, and those vary by maker, as above.

Do not confuse it with NVIDIA's actual wireless work: AI-RAN, computers for 6G mobile networks. NVIDIA is building these with Nokia, and announced a US$1 billion investment in the company in October 2025. Huang covered that in some of the same keynotes.

At GTC Washington, Huang described that telecom computer as able to "communicate wirelessly", so if you heard "wireless" there, that may be what was meant.

For what Rubin costs, how much faster it runs and when it reaches most buyers, read Vera Rubin price, performance and timing, or go back to the overview, Vera Rubin explained.

Where these figures come from

Figures are as of 24 October 2026. Ratios marked "our arithmetic" are calculated from these published figures.

Frequently asked questions

Does Vera Rubin have more memory than the B300?
No. Each Rubin GPU has 288 GB, the same as the B300, but it is faster HBM4. For an 8-GPU HGX Rubin NVL8, NVIDIA lists 2 TB at 130 TB/s: about 16 TB/s per GPU, or about twice the B300’s 8 TB/s. The HGX B300 has 2.1 TB. In the Vera Rubin NVL72 rack, NVIDIA lists 19.2 TB/s per GPU, about 2.4 times. So a model that fits on a B300 node should run faster on Rubin, but a model that does not fit will not fit on Rubin either.
How much power does Vera Rubin need?
Much more than Blackwell. NVIDIA lists about 24 kW for an 8-GPU DGX Rubin NVL8 server, against about 14 kW for a DGX B300. A Vera Rubin NVL72 rack needs 156 to 227 kW, depending on the power profile. Each Rubin GPU is rated at up to 2,300 W. Everything on the compute trays is liquid-cooled, with coolant at up to 45°C, so you need a liquid-cooled data centre hall.

Planning around Vera Rubin?

For most buyers Rubin matters from 2027. Register early interest and we will come back with pricing and availability as our partners confirm them. Need GPUs sooner? Tell us your GPU count, term and region and we will quote B300, B200 or GB300 NVL72 options.