Jensen just killed the CPU market. Nobody noticed.

There's a moment on every great earnings call where the CEO says something that sounds routine but is actually a declaration of war. Jensen Huang had one of those moments on Tuesday.

NVIDIA Corporation (NASDAQ:NVDA) posted $82 billion in revenue, up 85% year-over-year. The Q2 guide came in at $91 billion. These numbers are so large they've stopped feeling real. Analysts nodded. Markets moved. Pundits wrote their takes.

But buried inside that call was something far more important than the headline revenue: Jensen is going after the CPU market.

Almost $20 billion in standalone Vera CPU visibility this year alone. That's not a rounding error. That's NVIDIA entering a market it has never competed in — and doing it at a scale that Intel (NASDAQ:INTC) and AMD (NASDAQ:AMD) are not remotely prepared for.

Here's the chain reaction that most commentators missed entirely.

AI-native cloud companies, sovereign nations building their own compute infrastructure, and enterprises that won't touch public cloud — these are exactly the customers AMD's EPYC chips and Intel's Xeon processors have been counting on as their lifeline during the AI transition. They're the fallback market. The "at least we'll win there" market. And Jensen just walked in and started writing purchase orders.

If Vera takes $20 billion in CPU revenue this year at that second tier of customers, AMD doesn't just lose a contract. It loses the narrative that EPYC is the smart alternative to NVIDIA's GPU empire. The customer who buys Vera CPUs buys them because they're already running NVIDIA's full stack. The CPU follows the GPU. AMD never even gets a meeting.

The Anthropic partnership sharpens this further. Anthropic was effectively outside NVIDIA's direct compute story until recently. Now Jensen is deploying their infrastructure across AWS, Azure, CoreWeave, SpaceX's xAI, and more at the same time. That's not a chip transaction. That's Anthropic — one of the two most important AI labs on Earth — becoming architecturally dependent on NVIDIA across every cloud it touches.

Think about what that means three years out. If Anthropic's models run on NVIDIA infrastructure everywhere, and Anthropic's API is embedded in thousands of enterprise applications, then every one of those applications has NVIDIA baked into its cost structure. You can't rip that out. You'd have to rewrite everything.

Jensen is also articulating physical AI as a third leg — robotics, autonomous systems, industrial deployments. The customers there aren't hyperscalers. They're manufacturers, logistics companies, defence contractors. These are CPU buyers. Or they were.

One small but revealing detail from the call: Colette Kress (NVDA's CFO) misquoted the dividend figure live — said $0.20 per share when Jensen had to correct her to $0.25. Minor slip. But Jensen's pauses were most pronounced when he discussed competitive positioning and the ACIE segment. He wasn't uncertain. He was choosing words carefully — the tell of someone who knows exactly how much he's revealing.

Bull case: Vera hits $50 billion in annual revenue by 2028 and NVIDIA owns the full-stack AI compute layer from GPU to CPU across every tier of the market. Bear case: ARM-based alternatives from the hyperscalers themselves (Amazon Graviton, Google Axion, Microsoft Cobalt) commoditise the CPU layer before Vera can establish lock-in, and the $20 billion visibility number doesn't repeat.

The smart bears will point to that risk. They're not wrong to. But Jensen has earned the benefit of the doubt on market entries.