Key Highlights
- Nvidia’s innovative Vera CPU has transitioned from development to active production, with initial shipments reaching Anthropic, OpenAI, Oracle Cloud Infrastructure, and SpaceXAI.
- Ian Buck, Nvidia’s VP, personally transported the inaugural units, beginning with Anthropic’s San Francisco location.
- The Vera processor boasts 88 specialized Olympus cores, memory bandwidth of 1.2 TB/s, and delivers 50% enhanced per-core performance versus conventional CPU architectures.
- Oracle Cloud Infrastructure has committed to deploying hundreds of thousands of Vera processors starting in 2026, marking the first hyperscale cloud adoption.
- SpaceXAI is conducting evaluations of Vera for reinforcement learning applications, with Elon Musk receiving a detailed system walkthrough.
Nvidia (NVDA) has successfully transitioned its Vera CPU from the announcement phase into full-scale production, with initial deliveries reaching several prominent artificial intelligence organizations.
Ian Buck, Nvidia’s Vice President of Hyperscale and High-Performance Computing, personally oversaw the delivery of the inaugural systems to Anthropic, OpenAI, Oracle Cloud Infrastructure, and SpaceXAI during a two-day period last week.
The delivery tour commenced at Anthropic’s SoMa district headquarters in San Francisco, where James Bradbury, who oversees compute operations at Anthropic, accepted the delivery. Buck demonstrated the technology using an exposed Vera CPU motherboard, providing Bradbury with detailed insights into the chip’s distinctive features.
“Expanding compute capacity represents a critical catalyst for model advancement,” Bradbury stated. “Vera’s emergence as a valuable component of the ecosystem for addressing agentic workload challenges is genuinely exciting.”
The subsequent delivery occurred at OpenAI’s Mission Bay campus, where Sachin Katti, responsible for compute infrastructure, met with Buck on an exterior balcony area. During their discussion, Buck used a screwdriver to open the system casing, revealing the internal components.
Musk Receives Detailed Demonstration
The day’s final delivery took place at SpaceXAI’s Palo Alto facility. The Nvidia delegation provided Elon Musk with a comprehensive overview of the system’s design. Musk posed detailed inquiries regarding core configuration, memory architecture, and thermal management.
SpaceXAI is currently assessing Vera’s capabilities for reinforcement learning operations and agent-based simulation frameworks within its training infrastructure.
The following Monday, the delivery tour proceeded to Oracle’s AI Customer Excellence Center in Santa Clara, where Oracle Cloud Infrastructure’s product and customer success teams received hands-on demonstrations with the unpacked hardware.
“Beginning in 2026, OCI intends to implement hundreds of thousands of Nvidia Vera CPUs, as agentic AI requires consistent performance at unprecedented scale,” stated Karan Batta, OCI’s overall product management leader.
Oracle Cloud Infrastructure represents the inaugural cloud services provider to commit to Vera deployment at hyperscale levels.
Understanding Vera’s Functionality
Vera represents Nvidia’s inaugural custom-designed CPU, engineered explicitly for agentic AI workloads — scenarios where models perform actions beyond simple responses, including executing code, invoking tools, and maintaining extended context states.
While GPUs manage intensive computational tasks, the coordination infrastructure surrounding them — including tool invocation, data transfer, environment isolation, and information retrieval — relies on CPU processing. Vera was specifically designed to excel in these operations.
The processor incorporates 88 custom Olympus cores, provides 1.2 TB/s memory bandwidth capacity, and achieves 50% superior per-core performance under maximum load when compared to traditional CPU designs.
“When AI models receive queries, the solutions frequently aren’t readily available,” Buck explained during the Oracle visit. “The models must actually create Python code to derive accurate responses. This explains the explosive growth in CPU demand.”
Vera additionally functions as the host processor within the Vera Rubin NVL72 configuration, integrating with Rubin GPUs through second-generation NVLink-C2C technology in a unified memory framework. According to Nvidia, this configuration operates at twice the energy efficiency compared to traditional infrastructure approaches.
Jensen Huang initially unveiled Vera at the GTC San Jose conference in March, characterizing it as Nvidia’s forthcoming multi-billion dollar revenue stream.
Get 3 Free Stock Ebooks
Discover top-performing stocks in AI, Crypto, and Technology with expert analysis.
- Top 10 AI Stocks - Leading AI companies
- Top 10 Crypto Stocks - Blockchain leaders
- Top 10 Tech Stocks - Tech giants
