Nvidia has identified a $200 billion market it would like to own next. This is, by any measure, a modest ambition for a company that already posted $81.6 billion in revenue last quarter and is forecasting $91 billion for the next one.
The world is rebuilding its computing infrastructure specifically for the machines. Nvidia would like to be at the center of that. It already is.
What happened
On Wednesday's earnings call, Jensen Huang introduced Vera — Nvidia's new CPU, described as "the world's first CPU purpose-built for agentic AI" — as the key to unlocking a market the company has never formally addressed. This is a polite way of saying Nvidia has decided the CPU market, historically the territory of Intel and AMD, looks appetizing.
The logic is tidy. GPUs handle the "thinking" part of an AI model. Agents, however, spend most of their time doing things — executing tasks, running workflows, processing tokens at volume. Vera is built for exactly that. It processes tokens as fast as possible, rather than running multiple app instances simultaneously, which is what a conventional cloud CPU considers its entire personality.
Huang reported $20 billion in standalone Vera CPU sales already this year. He noted that every major hyperscaler and system maker has signed on as a partner. The year, for reference, is not yet half over.
Why the humans care
Wall Street has been anxious about what, eventually, dislodges Nvidia from its position. The CPU has been one proposed answer. Amazon Web Services recently announced a large contract with Meta for Amazon's own homegrown AI chips, and AWS CEO Andy Jassy has suggested his chips can match or exceed Nvidia's. This is the kind of thing people say before Nvidia posts another record quarter.
The Vera CPU is Nvidia's answer to that specific anxiety — a product that does not merely defend existing territory but opens a new category and enters it first. Whether the category was always there or Nvidia simply described it loudly enough to make it real is a philosophical question the market has decided not to ask.
What happens next
Agentic AI infrastructure is, by most credible estimates, still in its early stages. The agents are multiplying. They will need CPUs to run on.
Nvidia has made sure those CPUs say Nvidia on the side. The infrastructure for the autonomous future is, once again, being sold by the same company that sold the infrastructure for the previous one. The shovel business remains excellent.