Nvidia has announced that Vera Rubin is now in full production, and OpenAI is planned to scale its deployment in the third quarter.

      Nvidia announced on Monday that its Vera Rubin platform is now in full production. Ian Buck, the company’s vice president of accelerated computing, informed reporters at Nvidia's headquarters that systems are currently being shipped to customers such as OpenAI, CoreWeave, Google Cloud, Microsoft Azure, Meta, and Dell. According to Bloomberg, OpenAI intends to implement Vera Rubin on a large scale in the third quarter.

      CoreWeave, among the first cloud providers to receive the new hardware, reported to Bloomberg that its NVL72 racks are achieving ten times the token output compared to the previous model. The NVL72 system features 72 Rubin GPUs paired with Vera CPUs and utilizes liquid cooling to streamline internal cabling, a design innovation showcased at the headquarters event. Buck explained that this cooling method allows for the removal of copper connections that previously restricted the packing density of components.

      The Vera CPU was developed by Nvidia to replace processors sourced from other manufacturers. During the briefing, Nvidia drew direct comparisons with AMD, asserting that the Vera CPU is almost twice as fast as AMD’s Turin chip on Python workloads, a benchmark that is significant due to Python's prevalence in AI inference software. Early adopters of the processor include Anthropic, OpenAI, Perplexity, SpaceX, and Oracle.

      This production milestone coincides with a challenging time for Nvidia's stock performance. Although the company’s shares have increased by nine percent this year, the overall chip index has surged by 66 percent in the same timeframe, with Intel, ARM, and AMD each more than doubling.

      Analysts predict that Nvidia’s revenue will rise by 82 percent to approximately $393 billion for the fiscal year; however, this growth rate has not resulted in the same level of share price momentum seen during the Blackwell cycle.

      Over the past two months, Nvidia has steadily developed the narrative surrounding Vera Rubin. Jensen Huang announced the platform's full production at Computex in early June, identifying Anthropic, OpenAI, SpaceX, and Oracle as initial recipients. The latest briefing provided performance metrics and customer endorsements that were not available during the keynote address, converting claims into measurable benchmarks.

      It is important to note that the performance figures highlighted by Nvidia are derived from its own assessments and those of its customers, rather than independent evaluations. The tenfold token increase reported by CoreWeave and the Python benchmark against AMD stem from the company's own announcements, not from third-party sources. Volume shipments to all mentioned customers and independent verification of these performance claims are still pending.