Jensen Huang just told Wall Street he’s found a brand new $200 billion market for Nvidia, and it’s not in GPUs. Speaking on Wednesday’s earnings call, the Nvidia CEO pinned that number on Vera, the company’s new CPU built specifically for agentic AI, according to TechCrunch AI. The pitch landed alongside another record quarter: $81.6 billion in revenue, with $91 billion forecast for the next.
Huang called Vera “the world’s first CPU, purpose-built for agentic AI” and said every major hyperscaler and system maker is already partnering to deploy it. TechCrunch AI reports Nvidia has already moved $20 billion worth of standalone Vera CPUs this year. That’s a serious opening number for a product introduced back in March.
Why a CPU Play Matters Now
Nvidia owns the GPU market. CPUs have always belonged to Intel and AMD, with Amazon’s homegrown silicon now muscling in too. Last month AWS landed a massive contract with Meta for its own AI CPUs, and CEO Andy Jassy has been openly saying AWS can match or beat Nvidia on chips.
Huang’s counter is that agentic AI changes the rules. GPUs handle the “thinking” part of a model. Agents, on the other hand, mostly run on CPUs to execute tasks. Vera is engineered to process tokens as fast as possible, not to juggle multiple app instances across cores like classic cloud CPUs. Different workload, different chip.
The Agent Math Behind the $200B Claim
Huang’s logic is simple:
- The world has roughly a billion human PC users today.
- Soon it’ll have billions of AI agents.
- Each agent needs tools, basically its own PC-equivalent.
- More agents means a lot more CPUs.
Vera is sold standalone and bundled with the Rubin GPU. That bundling matters because it lets Nvidia pull customers who came for GPUs into a CPU relationship they didn’t have before.
What This Signals for the Industry
Three things stand out here.
- The agent economy is becoming a hardware story. Until now, most agent talk has been about frameworks, models, and orchestration. Huang is reframing it as a silicon buildout.
- Nvidia is no longer staying in its lane. A $200 billion TAM claim aimed directly at Intel, AMD, and Amazon’s chip ambitions is a declaration that Nvidia wants the full stack.
- Hyperscalers are hedging both ways. They’re building their own chips and buying Vera. That tells you nobody knows yet who wins the agentic CPU race, so they’re paying to be in every camp.
Huang has earned the benefit of the doubt by delivering quarter after quarter. Whether Vera actually unlocks $200 billion or settles somewhere smaller, the bigger signal is clear: the next phase of AI infrastructure spending is shifting from training GPUs to the chips that run agents in production.
Expect AWS, Google, and Microsoft to respond fast, and expect Intel and AMD to sharpen their own agentic CPU stories within the next two quarters. Full details at the original TechCrunch AI report.