NVIDIA's Vera Rubin nears $20B in its first shipping quarter
The new AI racks are running at CoreWeave, Google Cloud and others, but NVIDIA expects its gross margin to dip as it absorbs higher supply costs.

In short
- 01NVIDIA says its Rubin platform generated nearly $20B in its first quarter of production shipments, about a sixth of its $115.8B revenue.
- 02Each Vera Rubin NVL72 rack holds 72 Rubin GPUs and 36 Vera CPUs, and NVIDIA says racks are running at five named cloud partners.
- 03NVIDIA expects gross margin to fall from 74.8% to between 71% and 72% this quarter, a level its CFO calls the bottom.
The story
NVIDIA just confirmed Vera Rubin NVL72 has shipped, and its new AI platform is already a big business. On its earnings call transcript, the company said Rubin generated nearly $20B in its first quarter of production shipments.
That's about a sixth of NVIDIA's $115.8B in quarterly revenue, and the company calls it the fastest product ramp in its history. NVIDIA's supply commitments also rose to $390B from $279B in three months, CNBC reported.
Vera Rubin racks are already running at major clouds
Each NVL72 rack holds 72 Rubin GPUs and 36 Vera CPUs, built to work as one AI supercomputer. NVIDIA says racks are running at partners including CoreWeave, Google Cloud, Microsoft Azure, Oracle Cloud Infrastructure and Nebius.
Smaller clouds are getting them too. Nscale's launch release says its first Dell-built systems are online, with a wider rollout expected in the coming months. NVIDIA also says big cloud partners got Vera Rubin into production in about a month, half the time Grace Blackwell took.
The biggest speed claims still come from NVIDIA
In NVIDIA's MLPerf Inference post, a preview Vera Rubin system delivered up to 3.7x the throughput of GB300 NVL72 on Qwen3-VL. It reached up to 2.5x on DeepSeek-R1. MLPerf is an industry benchmark, but these are NVIDIA's own submissions.
NVIDIA's boldest figure is 10x more tokens per watt than GB300. Analyst Ian Cutress of More Than Moore cautioned that the number blends architecture, rack design and software maturity. He's still waiting for independent tests.
Higher supply costs will squeeze NVIDIA's margin this quarter
The ramp isn't free. NVIDIA's gross margin was 74.8% last quarter, and it expects 71% to 72% this quarter. CFO Colette Kress told analysts that's the low point.
What we provided here in Q4 is our bottom.
Colette Kress, NVIDIA CFO, on the earnings call
Cutress wrote that NVIDIA took on higher supply costs, memory in particular, while honoring orders placed at older prices. NVIDIA expects margins to recover to between 72% and 73% in fiscal 2028.
Customers can already rent the hardware, with general availability on Oracle Cloud Infrastructure and Google Cloud's A5X instances. NVIDIA forecasts $130B in revenue this quarter, plus or minus 3%, helped by an extra week. It says Vera Rubin Ultra is on track for next year.
Why it matters
Vera Rubin shows how quickly NVIDIA can move a new rack design into customers' data centers, and that speed is hard for rivals to match. The catch is cost, since even NVIDIA is giving up margin to absorb pricier supplies, memory included. Independent tests will show whether the performance gains are worth what buyers are paying.
Share


