The number that mattered walked in a mile ahead of expectations

Nvidia didn't just beat Wall Street on Wednesday, it beat by a margin that made the beat itself the headline. The company reported fiscal second-quarter revenue of $96.2 billion for the period ended July 26, up 106% year over year and well ahead of the roughly $92.2 billion analysts had penciled in. Non-GAAP earnings came in at $2.22 per share against expectations clustered around $2.06 to $2.09. For a company whose own guidance three months earlier had called for revenue near $91 billion, plus or minus 2%, actually landing at $96.2 billion isn't a modest overshoot. It's the kind of gap that tends to reset how the market prices the next several quarters, not just the one that just closed.

CEO Jensen Huang framed the results in a statement that read less like standard earnings-call boilerplate and more like a thesis statement for where he thinks the AI buildout stands. "AI has reached its inflection point," Huang said. "It's doing useful work. Its tokens are productive and profitable. Now, compute is revenue." He added, simply, that demand is accelerating.

Where the growth is actually coming from

The headline revenue number tells only part of the story; the composition underneath it is where the real signal sits. Data center revenue, still the overwhelming majority of Nvidia's business, came in around $89 billion against analyst estimates near $85.7 billion. Nvidia has also started breaking that segment into finer categories, and the split matters: revenue from its AI Clouds, Industrial, and Enterprise category, what the company calls ACIE, hit $40.3 billion, up 138% year over year, while hyperscaler revenue more than doubled compared to the same period last year. That's a meaningfully broader customer base than the narrative of "Nvidia only sells to five giant cloud companies" would suggest, even as those five companies remain enormous individual customers.

Gross margin held at 75% for a second straight quarter, an unusually high figure for a hardware business of this scale, though Nvidia guided that margin down slightly to 74% for the current quarter, a small signal that component costs, particularly in memory, are starting to bite even a company with Nvidia's negotiating leverage.

The guidance is the real story

Beating a quarter that's already happened is one thing. What actually moved the stock conversation Wednesday night was the forward number: Nvidia guided next quarter's revenue to $108 billion, plus or minus 2%, roughly $3.8 billion above the $104.2 billion Wall Street had modeled. And Nvidia delivered that guidance while explicitly stating it includes zero data center sales to China, a market the company has effectively been locked out of for its most advanced chips under existing export restrictions. A guidance beat that size, arrived at without assuming any recovery in a major market that's currently closed off entirely, is a stronger statement about underlying demand than a guidance beat that quietly assumes a policy tailwind nobody has actually confirmed yet.

That framing matters for how investors should read the number. It means the $108 billion figure is essentially Nvidia telling the market: this is what demand looks like even in the worst-case regulatory scenario for our second-largest addressable market. Any actual loosening of China restrictions would represent pure upside on top of an already aggressive number, a dynamic that echoes how Nvidia's other big strategic bets, including its stake tied to Musk's SpaceX and xAI, have quietly compounded well past their original scope.

The Amazon deal buried in the release

Alongside the earnings, Nvidia and Amazon announced that Amazon Web Services will purchase 2 million Nvidia GPUs, paired with Nvidia's new Vera CPU, with some units integrated directly into Rubin, Nvidia's forthcoming next-generation AI chip platform, and others deployed standalone. That's a genuinely large single commitment even by the standards of an industry that's grown used to multibillion-dollar infrastructure announcements, and it lands at a moment when investors have spent months quietly worrying that hyperscaler capital spending might finally be nearing a ceiling.

Nvidia CFO Colette Kress addressed that worry directly on the earnings call, saying capital expenditure among the "top five hyperscalers" is projected to climb to $1.3 trillion next year, up from roughly $800 billion in 2026. That's not a company managing decelerating demand; it's a company describing customers who are still increasing their spending commitments by more than 60% year over year, even after several years of already-enormous buildout.

What this means heading into a jittery week for markets

The timing here is not incidental. Nvidia's results landed in the same week Kevin Warsh delivers his first Jackson Hole address as Federal Reserve Chair, a speech markets are already parsing closely for signals on the September rate decision. A blowout Nvidia quarter, arriving right before that address, reinforces one side of an argument playing out inside the Fed itself: that AI-driven productivity gains are real and large enough to matter for the broader economy, not just for chip stocks. Whether that argument wins out in how the Fed actually sets policy remains genuinely unresolved. But Wednesday's numbers make it considerably harder to dismiss the AI investment cycle as running out of runway, at least for now, and considerably easier for anyone betting the other side of that trade to feel uneasy heading into Friday.