Delivering Vera: NVIDIA’s First CPU Built for Agents Is Shipping Now

Nvidia says Vera systems have begun shipping to AWS, after deliveries to Oracle Cloud Infrastructure, Anthropic, OpenAI and SpaceXAI. The CPU has 88 custom Olympus cores, 1.2TB/s memory bandwidth and up to 1.8x faster per-core performance on agentic workloads, according to Nvidia.

OCI plans to deploy hundreds of thousands of Vera CPUs beginning in 2026, while AWS plans to bring Vera-based infrastructure alongside a previously announced two-million-GPU expansion. Vera also hosts the CPU side of the Vera Rubin NVL72 system.

The pitch is that agents create CPU-heavy work through tool calls, sandboxes, retrieval and orchestration. If that workload grows, a purpose-built host CPU can prevent expensive GPUs from waiting on control and data movement. The source gives a clear mechanism, but adoption and independent evidence will decide its significance.

Signal Desk's read: The pitch is that agents create CPU-heavy work through tool calls, sandboxes, retrieval and orchestration. If that workload grows, a purpose-built host CPU can prevent expensive GPUs from waiting on control and data movement. The move is strategically coherent, although its limits are visible and should not be confused with proof of broad success.

This is a vendor-authored post; its claims are vendor-reported and need outside validation. The practical consequence is a new burden or option for operators, developers and institutions. That is where the announcement will be tested.

The source is an NVIDIA Blog post updated by Nvidia and includes partner statements; its specifications, efficiency claims and deployment narrative are vendor-reported.

Vera’s actual enterprise value will depend on end-to-end agent throughput and cost, not per-core figures. The post provides no independent comparison or production workload results.