Nvidia posts $81.6B revenue with 92% data-center growth, signals $200B push into server CPUs
Nvidia's print this week reframed the AI-infrastructure debate. Revenue of $81.6B beat the $78.9B consensus, EPS came in at $1.87, and — the figure that matters most for the AI build-out narrative — data center revenue grew 92% year over year. Jensen Huang's word for demand on the call was 'parabolic,' a deliberately stronger framing than the 'sustained' or 'unprecedented' language Nvidia has used in prior quarters. The 25x quarterly dividend hike (from 1¢ to 25¢) is a financial-signaling move pointed at long-only holders: management wants the cash-return story alongside the growth story.
The more strategically interesting disclosure is the server-CPU push. Analysts at The Motley Fool flagged a roughly $200B addressable opportunity Nvidia is now openly targeting — directly into the markets where AMD's EPYC and Intel's Xeon have lived. Nvidia's Grace and Vera-class Arm CPUs, tightly coupled to its GPU and networking stack, are now being marketed as the 'AI factory' substrate rather than as accessories. For AMD, which has been the AI-server-CPU winner of the last two years, this is the most direct threat yet; for Intel, it deepens an already painful competitive picture.
Nvidia is also continuing to clean up its consumer story. The GeForce 610.47 WHQL driver release drops support for the legacy Nvidia Control Panel, completing the transition to the App-centric driver UX. And Huang issued a separate warning to Super Micro and other ODMs around AI-chip smuggling scrutiny, signaling that export-control compliance is now a board-level supplier-management issue, not just a Washington headline.
The skeptical read: a 92% data-center growth rate is a deceleration from prior quarters by definition, and the 'parabolic' language is doing a lot of work to keep the multiple intact. Pair that with Source D's Energy News Beat report that Uber burned its full-year AI budget in roughly four months after Claude Code rollout, and the question hanging over Nvidia's print is whether end-customer AI economics can sustain the build-out Jensen is describing. For now, the market is voting yes.