Nvidia Q2 Earnings: $96B Revenue, $108B Guidance

Nvidia delivered the quarter Wall Street had spent weeks bracing for. After Wednesday’s close on August 26, 2026, the chipmaker reported fiscal second-quarter revenue of $96.2 billion, up 106% from a year ago and 18% from the prior quarter, and guided the current quarter to $108 billion — comfortably above the roughly $104 billion analysts expected. The print beat on every headline line and then raised the bar, the exact combination the market had feared it might not get after a jittery run into the report.

Shares dipped briefly in after-hours trading before turning higher, and Nasdaq 100 futures climbed as the guidance eased fears that the AI spending boom is losing steam. Nvidia had entered the print on a losing streak — eight down days out of nine — that reflected how much anxiety had built up around a single company’s results. As one framing on the tape put it afterward, the reaction showed Wall Street’s obsession with AI is far from over.

The headline numbers

For the quarter ended July 26, 2026, Nvidia reported:

• Revenue: $96.2 billion, up 106% year over year from $46.7 billion and up 18% sequentially.
• GAAP net income: $59.7 billion.
• GAAP earnings per share: $2.46; non-GAAP EPS: $2.22, up roughly 111% from $1.05 a year earlier.
• Gross margin near 75%, holding for a second straight quarter.

Analysts had penciled in something closer to $92 billion in revenue and non-GAAP EPS around $2.06 to $2.09, so the beat was decisive on both the top and bottom lines. It is the kind of quarter that has become routine for Nvidia and yet still lands as a shock in absolute terms: the company earned nearly $60 billion in profit in a single three-month stretch.

Data center is now the whole story

The Data Center segment generated $89.02 billion, up 117% year over year and now roughly 92% of total revenue. Gaming, professional visualization, and automotive still exist on the income statement, but Nvidia has effectively become an AI infrastructure company with a legacy graphics business attached.

Management attributed the acceleration to hyperscale customers ramping Blackwell Ultra deployments, with the Vera Rubin platform now in full production. CEO Jensen Huang framed the moment in unusually direct terms: “AI has reached its inflection point. It’s doing useful work. Its tokens are productive and profitable. Now, compute is revenue.” Demand, he said, “is accelerating” — a phrase carrying more weight than usual given how much of the market’s 2026 advance has rested on that single assumption. The dynamics behind that demand are the same ones driving the broader AI capex boom in hyperscaler spending.

The guidance that moved the stock

Nvidia guided fiscal third-quarter revenue to $108 billion, plus or minus 2% — implying year-over-year growth of more than 89% and topping the roughly $104 billion consensus. Even more striking was Huang’s forward commentary: he signaled expectations for something on the order of 70% revenue growth in fiscal 2028, well above what analysts had modeled for a company this size.

That outlook is what separated a good print from a market-moving one. Beating the current quarter matters, but investors have grown conditioned to Nvidia beats. What they wanted was evidence that the demand curve extends well into 2027 and beyond — that the orders behind today’s revenue are not a one-time build-out. The 70% figure, and the guidance above consensus, was the reassurance the tape was looking for after a nervous week previewed in our chip-stocks-into-Nvidia-earnings coverage.

Margins, memory, and the one soft spot

The blemish, to the extent there was one, sat in the margin outlook. Gross margin is guided to about 74% next quarter, down from roughly 75%, and management flagged that it expects margins to bottom in the 71% to 72% range in the fourth quarter of fiscal 2027 before recovering. The culprit is component cost inflation — above all memory prices, which have been climbing through the AI memory supercycle as high-bandwidth memory demand outstrips supply.

That pressure showed up elsewhere on the balance sheet, too. Nvidia’s supply commitments more than doubled, from $119 billion last quarter to $279 billion, driven largely by advance procurement of memory and other components. It is a striking number: the company is locking in a quarter-trillion dollars of future supply to guarantee it can build what customers are ordering. Bulls read it as visibility into demand; skeptics read it as a large fixed obligation if the cycle ever turns.

China, buybacks, and the cash machine

Nvidia’s outlook assumes no data center revenue from China, a continuation of the export-control overhang that has removed one of the company’s largest historical markets from the model entirely. That the company can guide to $108 billion with China zeroed out underscores how much demand exists everywhere else.

The cash generation remains staggering. Nvidia returned approximately $26.0 billion to shareholders during the quarter through buybacks and dividends, and still had about $99.0 billion remaining under its repurchase authorization at quarter’s end. It declared a quarterly dividend of $0.25 per share, payable October 1 to holders of record on September 10. A company producing this much free cash flow can fund its own supply commitments, buy back stock, and pay a dividend without straining — a luxury few in the AI hardware complex share.

What it means

Nvidia’s quarter functions as the closest thing the market has to a real-time gauge of the AI build-out, and the read this time was unambiguously strong. The combination of a clean beat, above-consensus guidance, and a 70% growth signal for fiscal 2028 tells investors that the orders underpinning the boom extend well past the current quarter — the single most important question hanging over the entire trade.

The winners are broad. Every hyperscaler committing capital to AI infrastructure gets validation that demand is real and durable, and the memory suppliers benefiting from Nvidia’s $279 billion in supply commitments gain a customer willing to pre-pay for capacity. The chip-equipment and networking names levered to data center build-outs ride the same wave. The clear message to the market is that “compute is revenue” is no longer a slogan but an income statement.

The risks are quieter but worth watching. The margin trough in the 71% to 72% range is a reminder that Nvidia’s suppliers — particularly memory makers — now hold real pricing power, and that input costs can erode even a dominant franchise’s economics at the edges. The $279 billion in supply commitments is a bet that demand holds; if the AI capex cycle ever cools, that obligation converts from a bull-case data point into a balance-sheet liability. And China remains a structural hole in the model, a market Nvidia has learned to grow without but would surely rather have back.

For now, the number that matters most is the one Huang put on next year: roughly 70% growth, on top of a business already running at a nearly $400 billion annualized revenue pace. As long as that trajectory holds, Nvidia remains the load-bearing wall of the entire market’s AI thesis. What to watch next is whether the hyperscalers’ own capex guidance — and the returns they can show on it — keep pace with the compute they keep ordering.