NVIDIA Begins Vera Rubin Shipments, Expects Fastest Product Launch in Its History
Key Facts
NVIDIA confirmed that it began production shipments of its Vera Rubin platform earlier in August, and said it expects the platform to mark the fastest product launch pace in the company’s history. On its fiscal 2027 second-quarter earnings call, it said it had already received purchase orders from all major hyperscale cloud providers, AI-focused cloud companies and original equipment manufacturers. The statement turns the news from an announcement that shipments have begun into a test of the company’s ability to convert stated demand into actual deployments. The platform is part of an expansion in infrastructure dedicated to AI workloads, at a time when integrated computing systems are coming into focus alongside processors themselves. The company did not provide a figure on the volume of shipments or the value of orders tied to Vera Rubin on the call.
Vera Rubin comprises five racks designed to work together as a single supercomputer for agentic AI tasks, according to NVIDIA’s announcement that the platform had entered full production. The company said performance on agent tasks at scale is 10 times that of the previous Grace Blackwell platform. Its rack-scale design means deploying the system requires working with an interconnected system rather than a standalone chip; the platform brings together processors, graphics units and networking in a single architecture. The company frames this performance in the context of running agent workloads at scale, while the figure remains a comparison provided by NVIDIA. These specifications provide technical context for why the announcement focused on AI “factories” and complete computing systems.
NVIDIA also presents the Vera Rubin POD as a larger architecture comprising 40 racks and 1,152 Rubin GPU units, with aggregate bandwidth for scaling of 10 petabytes per second. The specifications cited by the company include 1.2 quadrillion transistors, about 20,000 of its chips and computing capacity of 60 exaflops. The NVL72 rack integrates 72 Rubin GPU units and 36 Vera CPU units. The company says it has entered full production and is on track to ship in the second half of 2026. These different levels, from rack to POD, show how the platform is built to scale capacity across connected systems. The second half of 2026 remains the timeframe NVIDIA set for NVL72 shipments in its platform specifications.
On the customer front, NVIDIA said Meta plans to expand its deployment of Vera Rubin and Vera processors to support next-generation AI workloads. The collaboration between the two companies also includes Muse and its model family, according to NVIDIA’s announcement of an expanded partnership. This places Meta among the organizations with plans to deploy Vera Rubin infrastructure in partnership with the company, alongside collaboration related to models. The company did not say the platform was designed specifically to run Muse; the announcement establishes collaboration and expanded deployment, but does not specify an exclusive allocation of the platform to that model. NVIDIA and Oracle also announced on October 27 that Vera Rubin NVL72 and Vera processors were generally available through Oracle Cloud Infrastructure, as part of a plan to expand the infrastructure to a multigigawatt scale.
The move comes as NVIDIA’s business grows strongly. The company reported revenue of $96.2 billion in the second quarter of fiscal 2027, ended July 26, up 106% year over year and 18% from the previous quarter. Data center revenue was $89 billion, up 117% from the same period a year earlier. In August, the company forecast revenue of about $108 billion for the current quarter, running from August to October. But Chief Executive Jensen Huang said the “entire supply chain is facing challenges,” and that available supply covered 70% of demand. The chief financial officer added that revenue growth for the fiscal year ending in January 2028 could have approached a doubling if not for difficulties securing supply, making component availability a factor in converting demand into shipments and revenue.
After NVIDIA’s August earnings call, its shares rose 4.1% in after-hours trading, having fallen 1.6% in the regular session, according to the Associated Press. In the latest session listed in EL7 data, the shares closed at $202.06 on November 2, compared with $204.64 on October 30. These figures track the share price around the announcement and in subsequent sessions, but do not isolate Vera Rubin’s impact from other factors affecting trading. NVIDIA scheduled its fiscal 2027 third-quarter earnings call for November 17, 2026; the quarter ended October 25. The forthcoming results are expected to provide an update on quarterly revenue following the start of production shipments, and on what the company says about demand and supply.