Nvidia Ships Vera CPU to Challenge Intel and AMD in AI Servers

Nvidia has begun shipping its first fully custom-designed data center CPU, called Vera, to major AI customers including OpenAI, Anthropic, and SpaceX. The move positions the GPU leader as a direct competitor to Intel and AMD in the server processor market, according to details first reported by CNBC.

The company released full specifications and benchmarks for Vera on Tuesday, providing the technical documentation cloud providers need to evaluate the chip. Deliveries to initial customers occurred in June, Nvidia representatives confirmed.

Why it matters

The rise of agentic AI—systems that operate autonomously with minimal human oversight—has renewed focus on CPU performance. These agents require fast processors to manage tasks and feed data to GPUs efficiently. Nvidia's entry into the CPU market represents a strategic vertical integration play that could reshape data center economics and threaten the duopoly Intel and AMD have maintained for decades. With the server CPU market valued at approximately $37 billion in 2025 and Nvidia projecting the AI-optimized segment could reach $200 billion, the stakes extend well beyond chip sales to control of the AI infrastructure stack.

Designed for AI agents, not traditional workloads

Vera marks Nvidia's first server CPU designed from the ground up, rather than using off-the-shelf Arm architecture. The chip prioritizes single-core speed over core count, a departure from traditional server CPU design philosophy that Intel and AMD have followed.

"Particularly how fast a CPU can answer one question," Ian Buck, Nvidia's vice president of hyperscale, said at a presentation, explaining why CPUs have become "much more integral" in the age of AI agents.

Nvidia claims Vera delivers 50% better performance for AI agent workloads compared to x86 processors from Intel and AMD. Hannah Coutand, a Vera product marketer at Nvidia, explained the chip focuses on per-core speed, high memory bandwidth, and low latency "so that agents can return to their GPUs as quickly as possible and keep those GPUs, which are a very expensive and a highly valuable asset in the AI factory, as highly utilized as possible."

The processor supports up to 1.5 terabytes of low-power memory per chip and consumes between 250 and 450 watts of power.

Market dynamics and adoption challenges

Nvidia will sell Vera as a standalone chip and in various configurations, including a liquid-cooled rack containing 256 Vera processors and systems paired with Nvidia GPUs called Vera Rubin. Wolfe Research estimated an average selling price of approximately $5,000 per chip and projected shipments of 1.3 million units this year, though Nvidia declined to comment on pricing.

While OpenAI plans large-scale Vera deployments starting this quarter, adoption among major cloud providers remains in "early innings," according to Coutand. Oracle is currently the only major cloud service provider listed as a partner.

AMD holds roughly 33% of the server CPU market while Intel commands 66.8%, according to Gartner analyst Kevin Knox, who identified AMD as "the company to beat in enterprise AI server CPUs" due to strong hyperscaler relationships. Both companies have seen their stock prices surge in 2026—AMD up 128% and Intel up 149%—outpacing Nvidia's 8% gain as investors recognize the renewed importance of CPUs in AI infrastructure.

Cambrian AI Research founder Karl Freund noted that Nvidia has created a new CPU category without direct competition from Intel or AMD yet, but characterized the effort as a strategic move to "unhook their customers from using Intel or AMD CPUs" and capture that revenue stream.

The details were first reported by CNBC.