Jensen Huang is the corporate hype man we’ve come to expect. Every earnings call brings another sweeping proclamation about Nvidia’s future, delivered with the kind of conviction that makes Wall Street either nod along or brace for impact.
But here’s the thing: he usually backs it up.
So when Huang announced during Wednesday’s earnings call that he’s found a “brand new $200 billion TAM” for Nvidia, it’s worth taking seriously, even if skepticism feels like the natural instinct. The company just posted $81.6 billion in revenue and guided for $91 billion next quarter. At this point, Huang has earned at least the benefit of the doubt.
The bet centers on Vera, a CPU Nvidia introduced in March. Not a GPU, where Nvidia is essentially unbeatable. A CPU. The market that historically belonged to Intel and AMD, not the graphics chip giant.
The Agent Opportunity
Huang’s pitch is surprisingly logical, even if it sounds like science fiction. As AI systems evolve from “thinking” models that run on GPUs to autonomous agents that take actions in the world, those agents will need CPUs to execute tasks. Lots of them.
“The world has a billion users, human users,” Huang said on the call. “My sense is that the world is going to have billions of agents, and those billions of agents will all use tools. And those tools are going to be like PCs, just like us humans using PCs today.”
Vera isn’t a traditional CPU built around running multiple applications simultaneously. It’s purpose-built to process tokens as fast as possible, which is what agentic AI demands. Different architecture. Different optimization. Different value proposition.
And apparently, different market demand. Huang claims Nvidia has already sold $20 billion worth of Vera CPUs this year alone, bundled with its Rubin GPU or sold standalone. If that number is accurate, it suggests hyperscalers aren’t waiting around to see if this works.
But The Competition is Real
Here’s where the hype starts to strain credibility. Amazon Web Services, Meta, Google, and various startups are all developing their own AI chips. They’re not doing this as a vanity project. These are trillion-dollar companies that believe they can build better, cheaper, or more tailored silicon than Nvidia can provide.
Last month, AWS announced it had signed a massive contract with Meta for millions of its homegrown AI CPUs. Andy Jassy has been explicit: AWS thinks it can match or beat Nvidia at chip design, both GPUs and CPUs.
So what makes Huang think Nvidia will be the default choice for agentic CPUs when the most powerful cloud providers on Earth are actively building alternatives?
The answer, he suggests, is velocity and trust. Nvidia has already shipped billions of dollars worth of Vera. The hyperscalers are “partnering with us to deploy it,” according to Huang. They’re not waiting. They’re already building Vera into their infrastructure.
The Trillion-Dollar Question
There’s a real tension here worth acknowledging. Huang’s optimism about a new $200 billion market feels earned given Nvidia’s track record of execution. But the competitive dynamics in chip design are shifting. When your biggest customers are also your most capable competitors, claiming a monopoly on any new chip category gets harder to defend.
Vera could absolutely become the standard for agentic Technology stacks across the industry. Or it could become one option among many, with AWS, Google, and others carving out their own fiefdoms over time.
The difference between those outcomes isn’t billions. It’s hundreds of billions. And that’s probably why Wall Street remains anxious despite Nvidia’s dominance. Dominance breeds the incentive to compete.
