Nvidia just posted a quarter that would have sounded absurd a few years ago, and yet here we are. The chipmaker reported second-quarter revenue of $96.2 billion, more than double what it pulled in during the same period last year and roughly $4 billion ahead of what Wall Street analysts had penciled in. For a company that’s already the most valuable in the world, that kind of beat is hard to shrug off, and it says a lot about just how far hyperscalers are still willing to go to build out AI infrastructure.
The numbers behind the headline figure are almost as striking as the top line itself. Data center revenue alone came in at $89 billion, which now accounts for well over 90% of everything Nvidia sells. That’s the clearest sign yet that this is no longer really a “graphics card company” in any meaningful sense; it’s essentially the backbone supplier for the entire AI buildout happening across cloud providers, sovereign AI projects, and a growing list of enterprise customers. Adjusted earnings per share landed at $2.22, comfortably ahead of the roughly $2.09 consensus estimate, and more than double what Nvidia reported in the same quarter a year ago.
Looking ahead, Nvidia guided next quarter’s revenue to $108 billion, which is itself several billion dollars above what analysts were expecting. CEO Jensen Huang leaned into the momentum during the earnings call, declaring that “AI has reached its inflection point” and that the technology is now “doing useful work” with tokens that are “productive and profitable.” He framed the current moment as a shift from experimentation to genuine economic output, arguing that compute itself has effectively become revenue. According to Huang, the buildout is no longer being driven by a single dominant lab the way it was a year ago; instead, multiple frontier labs, a wave of new AI startups, and a growing open-model ecosystem are all scaling in parallel and adding to demand.
Why the Stock Still Slipped
Despite blowing past nearly every headline estimate, Nvidia shares wobbled in after-hours trading, and the reason comes down to a single number buried deeper in the report: margins. The company’s gross margin guidance for the upcoming quarter came in at 74%, a step down from the 75% it posted this quarter. Nvidia attributed the pressure to rising memory costs, a reminder that even a company sitting on this much pricing power isn’t fully insulated from supply chain dynamics further up the chain. For investors who’ve grown used to Nvidia clearing every bar set in front of it, even a modest one-point dip in margin guidance was enough to trigger some selling, even though the stock briefly swung higher immediately after the results dropped before settling into a more mixed reaction.
That kind of reaction has become something of a pattern for Nvidia lately. Despite beating earnings estimates comfortably in each of the last several quarters, the stock has tended to drift lower in the days that follow, as investors nitpick guidance details or worry aloud about how sustainable this pace of spending really is. It’s the classic high-expectations problem: when a company has trained the market to expect a blowout every single quarter, even genuinely excellent results can feel like a letdown if any single metric ticks in the wrong direction.
Betting Big on the Next Chip Generation
Nvidia’s forward-looking commentary also gave investors a clearer look at what’s coming next. Huang pointed to the ramp of the company’s Blackwell platform and the early production shipments of its next-generation Vera Rubin architecture, which he said is on track to surpass Blackwell and open up what he described as a $200 billion addressable market tied to CPU-related workloads. Notably, the Q3 revenue guidance assumes zero data center compute revenue coming from China, underscoring how much of Nvidia’s growth story is now being carried by demand elsewhere, even as geopolitical restrictions continue to keep the Chinese market largely off the table.
Huang also struck a confident note about supply, telling analysts the company has enough capacity lined up to support 70% year-over-year revenue growth heading into fiscal 2028. That’s a bold claim for a business already generating revenue at this scale, but it fits the broader narrative Nvidia has been pushing all year: that AI infrastructure spending isn’t a temporary spike but the early stage of a much longer buildout cycle.
For now, the takeaway from this quarter is a familiar one dressed up in bigger numbers. Nvidia keeps delivering results that would be considered spectacular for almost any other company on the planet, and the market keeps finding something to worry about anyway, whether it’s a one-point margin cut, memory costs, or questions about how long hyperscalers can keep spending at this rate. Jensen Huang, for his part, doesn’t seem worried. If his read on the AI inflection point holds up, this record quarter may end up looking modest compared to what’s still ahead.
