Media Investigation: Nvidia Rubin—Dozens of Customers Have Received Test Racks, Production Process Smoother Than Blackwell, Mass Production Expected Next Year

Nvidia's next-generation Vera Rubin server system is accelerating, with hopes for large-scale mass production by next year. AI developers, cloud service providers, and Wall Street investors are eagerly anticipating its arrival.

On July 22, according to The Information, several senior Nvidia executives recently gave in-person media tours, providing detailed disclosures on the progress of Vera Rubin’s mass production and technical details.

Andrew Bell, Nvidia’s Senior Vice President, stated that the manufacturing process of Vera Rubin is clearly superior to the previous generation Blackwell product, and he remains optimistic about the smoothness of this rollout.

The production cadence of Vera Rubin racks will have a decisive impact on Nvidia’s future revenues. At maximum capacity estimates, the potential revenue scale far exceeds Nvidia’s current quarterly performance, and the market’s attention to its commercial prospects continues to grow.

According to The Information, dozens of customers including CoreWeave, Microsoft, OpenAI, Anthropic, and SpaceXAI have already received a small number of Vera Rubin test racks, each equipped with 72 GPUs.

In terms of pricing, one buyer disclosed that each Vera Rubin rack is priced between $7 million and $8 million, whereas the current flagship Grace Blackwell 300 rack sells for about $5 million, showing a significant increase.

During the media’s visit to Nvidia’s Santa Clara headquarters, they also went to a semi-secret test site where about 30 server racks were running on-site, many of which were Vera Rubin. Nvidia executives stated that OpenAI is using some of these racks.

In terms of appearance, a Vera Rubin rack is about the size of a tall filing cabinet but weighs as much as 4,000 pounds, equivalent to a pickup truck.

Compared with the many setbacks encountered during the launch of Blackwell, Vera Rubin’s manufacturing progress is significantly smoother.

In an interview, Nvidia Senior Vice President Andrew Bell admitted that Blackwell had “almost everything collapse” — hardware issues, software issues, diagnostic system problems, and even manufacturing glitches, leading them to start over from scratch.

Bell stated that Vera Rubin features important design improvements: the racks contain almost no cables, and robotic systems handle most assembly steps, which will greatly reduce assembly difficulty and thus improve production efficiency and yield.

He remains cautiously optimistic that Vera Rubin can avoid the technical bottlenecks that Blackwell encountered.

Regarding production planning, Bell revealed that Nvidia's more than ten manufacturing partners will eventually be able to produce up to 1,000 Vera Rubin racks per day.

Based on this capacity, conservative estimates put potential single-quarter revenue at no less than $630 billion, while Nvidia’s actual revenue for the April quarter this year was $82 billion.

However, having maximum capacity does not mean actual deliveries will proceed at that rate—actual shipment volumes will still depend on customer demand and installation pace.

Nvidia Vice President Ian Buck said the company’s decisions on GPU allocation are directly tied to whether customers are truly physically ready to install servers, jokingly hinting that ultimate decision power lies with CEO Jensen Huang.