HomeTechHardwareNVIDIA Begins Shipping Vera CPUs for A
HARDWARE

NVIDIA Begins Shipping Vera CPUs for Agentic AI Workloads

NVIDIA has started delivering its 88-core Vera processors to AWS, Oracle Cloud Infrastructure, OpenAI, Anthropic, and SpaceXAI.

WHAT YOU NEED TO KNOW
  • NVIDIA has begun shipping Vera, its first custom CPU featuring 88 Olympus cores and 1.2TB/s of memory bandwidth.
  • AWS received its first Vera server and Rubin GPU under an expanded agreement that includes plans for two million additional GPUs.
  • Oracle Cloud Infrastructure plans to deploy hundreds of thousands of Vera CPUs starting in 2026.
  • Anthropic, OpenAI, and SpaceXAI received initial Vera systems to test agentic, simulation, and reinforcement learning pipelines.

NVIDIA has begun shipping its Vera CPU at scale, delivering initial hardware to Amazon Web Services, Oracle Cloud Infrastructure, Anthropic, OpenAI, and SpaceXAI, NVIDIA announced on Thursday. The processor represents the company's first custom CPU designed specifically for agentic artificial intelligence workloads.

NVIDIA Vice President of Hyperscale and HPC Ian Buck delivered the first Vera CPU server and Vera Rubin GPU to AWS headquarters in Seattle, handing the system to Amazon EC2 executives Willem Visser and Supreeth Sheshadri. The handoff follows an expansion of AWS and NVIDIA's partnership, which includes plans for two million additional NVIDIA GPUs and Vera-based cloud infrastructure.

The Vera processor includes 88 custom NVIDIA-designed Olympus cores and 1.2TB/s of memory bandwidth. NVIDIA stated the chip provides up to 1.8 times faster per-core performance on agentic workloads, which handle orchestration layers, code generation, tool calls, sandboxes, and long-context retrieval operations.

Cloud and lab deployments

Initial system deliveries began in May 2026 across cloud providers and artificial intelligence laboratories. Oracle Cloud Infrastructure plans to deploy hundreds of thousands of Vera CPUs starting in 2026, making it the first cloud provider deploying the hardware at hyperscale.

Research laboratories are testing the chips for specialized training stacks. SpaceXAI is evaluating Vera for reinforcement learning workloads and agent-based simulation pipelines in Palo Alto. OpenAI head of compute infrastructure Sachin Katti accepted a system in San Francisco, and Anthropic compute lead James Bradbury received a server walkthrough at the firm's SoMa offices.

System architecture

Vera functions both in standalone CPU servers and as the host processor for the Vera Rubin NVL72 system. In NVL72 configurations, Vera connects to a pair of Rubin GPUs across a second-generation NVLink-C2C interconnect, using a unified memory architecture alongside BlueField-4 DPUs and Spectrum-X networking to handle data movement at twice the energy efficiency of traditional infrastructure.

Xentir Media
Xentir Media NewsroomSource-backed AI and technology coverage, drafted by Xentir's automated editorial system under fixed human-set rules. See our editorial policy and AI usage policy.
J
Jomon · Founder & EditorFounder and editor of Xentir Media. Sets the editorial rules the newsroom system runs under, and is accountable for its corrections. About Jomon · [email protected]
The Xentir Brief
The developments worth knowing — one useful email.
Get the Brief →