NVIDIA Says Vera CPU Systems Are Shipping, With First Server Delivered to AWS

NVIDIA says systems built around Vera, its first custom CPU, are shipping and that Amazon Web Services has received a Vera CPU server and a Vera Rubin GPU. The handoff moves Vera from an announced product to delivered hardware at a major cloud provider.

It does not establish customer availability. Neither NVIDIA’s delivery report nor the related AWS announcements names a Vera instance type, region, price, ordering path or customer launch date. None says AWS customers can currently rent, reserve or buy Vera-based capacity.

NVIDIA’s August 27 delivery report says Vera systems have begun shipping and describes the first NVIDIA Vera CPU server, together with a Vera Rubin GPU, arriving at AWS in Seattle. The report does not say whether the hardware is an evaluation unit, a production deployment or part of a customer-facing fleet.

Vera is the first CPU NVIDIA has designed itself. The company introduced it on May 31 as a processor built for agentic AI: software that runs extended sequences of model calls, tool use and data handling rather than handling a single prompt and response.

NVIDIA says Vera has 88 custom Olympus cores and 1.2 TB/s of memory bandwidth, the rate at which the processor can move data to and from memory. It claims up to 1.8 times faster per-core performance on agentic-AI workloads. Those figures are NVIDIA’s; the benchmark configurations and comparison baselines have not been independently validated.

The company says Vera will serve as a standalone CPU server, the host processor in Vera Rubin systems and a component of Vera BlueField-4 STX infrastructure. That broadens NVIDIA’s AI infrastructure stack beyond the GPUs for which it is best known, but does not by itself show how cloud providers will package or sell the systems.

The delivery followed an expanded AWS and NVIDIA collaboration announced on August 26 that included work to bring Vera CPU-based infrastructure to AWS. The announcement set no instance name, region, price or availability date for Vera. It comes as AWS is also expanding its broader agentic-AI work with customers, a separate initiative that does not establish Vera availability.

“Customers want the freedom to choose the best tools for their AI workloads, and they want confidence that everything works seamlessly together,” AWS CEO Matt Garman said in the August 26 announcement.

AWS’s AI Factories FAQ separates NVIDIA accelerator options available today from planned offerings. It says AWS intends to offer Rubin GPUs and the Vera Rubin platform once they are generally available, subject to timing and configuration requirements.

In NVIDIA’s delivery report, Anthropic head of compute James Bradbury said, “We’re excited to see Vera emerge as a promising part of the ecosystem when solving for agentic workloads.”

When NVIDIA announced Vera in May, it said the chip was in full production and that systems would be available from system builders and cloud partners starting in fall 2026.