Nvidia (NVDA) Begins Shipping Vera CPUs to AI Giants Including OpenAI and Anthropic

Key Highlights

• Nvidia’s Vera CPU has transitioned from development to active production, with initial shipments reaching leading AI organizations like Anthropic, OpenAI, Oracle Cloud Infrastructure, and SpaceXAI.
• Ian Buck, Nvidia’s VP of Hyperscale and HPC, conducted personal deliveries to each company’s facilities.
• The Vera processor incorporates 88 Olympus cores with custom architecture, 1.2 TB/s memory bandwidth capabilities, and achieves 50% improved per-core speed compared to conventional CPUs.
• Oracle’s cloud division is preparing for a massive Vera deployment targeting hundreds of thousands of units starting in 2026, marking the first hyperscale cloud adoption.
• Elon Musk received a personal demonstration at SpaceXAI, where the company is exploring Vera for reinforcement learning applications.

Nvidia (NVDA) has successfully transitioned its Vera CPU from the announcement phase into full-scale production, with first deliveries reaching several heavyweight AI companies.

NVIDIA Corporation, NVDA

Ian Buck, who serves as Nvidia’s VP of Hyperscale and HPC, made the rounds last week delivering inaugural units directly to Anthropic, OpenAI, Oracle Cloud Infrastructure, and SpaceXAI across two action-packed days.

Buck’s initial destination was Anthropic’s South of Market district headquarters in San Francisco, where James Bradbury, the company’s compute lead, accepted delivery. Buck came prepared with an exposed Vera CPU motherboard to provide Bradbury with a detailed technical walkthrough.

Following Anthropic, Buck headed to OpenAI’s Mission Bay campus, where Sachin Katti, who oversees compute infrastructure, met him outside. In a hands-on demonstration, Buck even grabbed a screwdriver during their discussion to open the chassis and reveal the internal components.

Musk Receives Personal Demonstration

The day’s concluding stop brought Buck to SpaceXAI’s Palo Alto location. Nvidia’s delegation provided Elon Musk with a comprehensive system architecture briefing. Musk engaged with detailed inquiries covering core configuration, memory design, and thermal management.

SpaceXAI is currently testing Vera’s capabilities for reinforcement learning tasks and agent-based simulation frameworks within their training infrastructure.

Monday saw the tour extend southward to Oracle’s AI Customer Excellence Center in Santa Clara, where OCI’s product and customer success teams examined the unpacked system firsthand.

OCI stands as the inaugural cloud service provider committing to Vera at hyperscale proportions.

Understanding Vera’s Purpose

Vera represents Nvidia’s inaugural custom CPU design, engineered specifically for agentic AI workloads—scenarios where models move beyond simple query responses to execute actions, run code, invoke tools, and maintain extended context states.

While GPUs manage intensive computation, the surrounding orchestration infrastructure—including tool invocation, data transfer, sandboxing, and retrieval operations—requires CPU processing power. Vera was purpose-built for these tasks.

The processor features 88 custom Olympus cores, delivers 1.2 TB/s memory bandwidth, and achieves 50% superior per-core performance under maximum load relative to conventional architectures.

Vera additionally functions as the host processor in Nvidia’s Vera Rubin NVL72 configuration, connecting with Rubin GPUs through second-generation NVLink-C2C in a unified memory framework. Nvidia reports the configuration operates at double the energy efficiency of conventional setups.

Jensen Huang unveiled Vera at GTC San Jose in March, characterizing it as Nvidia’s forthcoming multi-billion dollar revenue stream.

The post Nvidia (NVDA) Begins Shipping Vera CPUs to AI Giants Including OpenAI and Anthropic appeared first on Blockonomi.