Nvidia’s Jensen Huang unveils Vera CPU designed for AI agents

Nvidia just told the world it’s not content being the king of GPUs. It wants the CPU crown too, at least when it comes to AI.

CEO Jensen Huang took the stage at GTC Taipei to introduce the Vera CPU, an 88-core custom Arm-based processor that Nvidia claims is the first chip designed explicitly for agentic AI workloads. Not general computing. Not traditional server tasks. Specifically the kind of work that autonomous AI agents do: executing code, calling tools, processing data, and making decisions without a human babysitting every step.

What Vera actually does differently

Nvidia says the Vera CPU handles agentic workloads 1.8 times faster than standard x86 processors, with up to 6x gains on specific metrics.

The Vera CPU sits within the broader Vera Rubin platform, which Nvidia introduced earlier in March 2026. That platform bundles Rubin GPUs alongside infrastructure components purpose-built for hyperscale agentic AI deployments.

First units have already shipped to some of the biggest names in AI. OpenAI, Anthropic, and SpaceXAI have all taken delivery, with initial deployments starting in mid-May 2026. Full-scale shipments are expected in the fall.

The $200 billion bet on agentic AI

Nvidia isn’t making a small wager here. The company estimates that the agentic AI computing market represents a $200 billion opportunity. And it expects to capture $20 billion in standalone CPU revenue this year alone.

The strategic logic is straightforward. As AI agents proliferate, they need both GPUs for inference and training and CPUs for orchestration, scheduling, and tool execution. By owning both sides of that equation, Nvidia can sell complete platforms rather than individual components.

What this means for investors

The more immediate investment angle is Nvidia itself. The company is essentially building the picks and shovels for what it believes will be a $200 billion gold rush. If agentic AI adoption follows the trajectory Nvidia is banking on, the CPU business alone could become a meaningful percentage of the company’s total revenue within a few years.

Investors watching the AI infrastructure buildout should track Vera Rubin platform adoption rates closely as full-scale shipments begin this fall, because early traction with tier-one AI labs tends to cascade into broader enterprise adoption within 12 to 18 months.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.