NVIDIA Corp's next-generation Vera Rubin server system is advancing rapidly, with expectations set for large-scale production next year, a development eagerly awaited by AI developers, cloud service providers, and Wall Street investors.
Recent media reports detail that several NVIDIA Corp executives provided an on-site look, sharing updates on the Vera Rubin system's production timeline and technical specifications.
NVIDIA Corp Senior Vice President Andrew Bell indicated that the manufacturing process for Vera Rubin is proceeding more smoothly than that of the prior Blackwell generation, expressing optimism about the progress.
The pace of Vera Rubin rack production is poised to be a decisive factor for NVIDIA Corp's future revenue. Based on maximum capacity estimates, the potential revenue scale significantly surpasses the company's current quarterly results, intensifying market focus on its commercial prospects.
Multiple Clients Have Received Test Racks
Reports indicate that dozens of clients, including CoreWeave, Microsoft, OpenAI, Anthropic, and SpaceXAI, have already received initial small quantities of Vera Rubin test racks, with each rack equipped with 72 GPUs.
Regarding pricing, a buyer disclosed that each Vera Rubin rack is priced between $7 million and $8 million. This represents a notable increase compared to the current flagship Grace Blackwell 300 rack, which is priced around $5 million.
During a visit to a semi-confidential testing site near NVIDIA Corp's Santa Clara headquarters, journalists observed approximately 30 server racks in operation, many of which were Vera Rubin units. Executives noted that OpenAI is utilizing some of these racks.
Physically, a Vera Rubin rack is roughly the size of a tall filing cabinet but weighs about 4,000 pounds, comparable to the weight of a pickup truck.
Production Process More Efficient Than Blackwell
Compared to the significant challenges encountered during the Blackwell launch, the manufacturing progress for Vera Rubin has been markedly smoother.
In an interview, Senior Vice President Andrew Bell candidly stated that with Blackwell, "almost everything broke"—there were hardware issues, software issues, diagnostic system problems, and manufacturing complications, ultimately necessitating a restart.
Bell explained that Vera Rubin incorporates key design improvements: the racks are nearly cable-free, and robotic systems handle a large portion of the assembly. These changes are expected to significantly reduce assembly complexity, thereby boosting production efficiency and yield rates.
He expressed cautious optimism that Vera Rubin will avoid the technical bottlenecks faced by its predecessor.
Maximum Daily Capacity and Revenue Potential
On the capacity front, Bell revealed that NVIDIA Corp's more than ten manufacturing partners are ultimately expected to achieve a combined capacity of up to 1,000 Vera Rubin racks per day.
A rough calculation based on this maximum capacity suggests a potential quarterly revenue exceeding $630 billion. For context, NVIDIA Corp's actual revenue for the quarter ending in April was $82 billion.
However, having maximum capacity does not guarantee shipments will proceed at that rate; actual volume will depend on customer demand and installation readiness.
NVIDIA Corp Vice President Ian Buck noted that the company's GPU allocation decisions are directly tied to whether customers are physically prepared to install the servers. He quipped that the final allocation decision ultimately rests with CEO Jensen Huang.
Comments