The stock dropped in the first minutes after the close, then turned around and climbed. Nvidia Corp. reported revenue of $96.2 billion for the quarter ended July 26, up 106 percent from a year earlier, and promised more growth ahead than Wall Street had modeled. Shares rose 7.63 percent in overnight trading, the kind of whipsaw that has become routine for the most valuable company in the world.
Revenue of $96.221 billion topped analyst expectations by more than 4 percent, the company said. Net income rose 126 percent to $59.688 billion, and adjusted earnings of $2.22 per share were up 120 percent. The data center segment, the engine of the company’s growth, brought in $89.0 billion, up 117 percent and more than nine-tenths of total revenue. Hyperscalers bought $48.71 billion of that; AI clouds, industrial, and enterprise customers accounted for the rest, with that slice growing 138 percent. Gross margin held at 75 percent, roughly flat with the prior quarter.
The guidance did the heavy lifting. Nvidia said third-quarter revenue would come in around $108 billion, plus or minus 2 percent, and that the figure assumes no data center compute revenue from China, where export and import licensing remains unsettled. The non-GAAP gross margin outlook of 74 percent, plus or minus half a point, came in below the 75 percent analysts had expected, a modest concession to the cost of ramping new products. Chief Financial Officer Colette Kress said the coming Vera Rubin platform would account for about 20 percent of data center revenue in the third quarter, roughly $20 billion in a single quarter, and called it the fastest product ramp in company history.
The longer view was what moved the shares. Management, for the first time, gave a growth target for the fiscal year that starts in January: revenue up about 70 percent. The street had been modeling roughly 45 percent. Jensen Huang, the chief executive, said AI had reached “a tipping point” and described demand that is still expanding faster than supply. Memory purchase commitments have climbed to $160 billion, up from earlier disclosed levels, as the company locks in the high-bandwidth components that gate its chip production.
The numbers suggest an industry still in the middle of a capital-spending cycle, not the end of one. Amazon’s cloud unit used the call to announce another two million Nvidia GPUs on order, spanning the Blackwell Ultra, Rubin, and Rubin Ultra generations, with delivery through 2028. Nvidia separately disclosed land, power, and shell guarantees for AI cloud partners totaling up to $108.5 billion in maximum gross exposure, including credit support tied to a 20-year lease at a 4.25-gigawatt campus in Ohio that will host OpenAI’s compute. It also said it had formed partnerships with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR to create financing platforms that could mobilize more than $500 billion for AI infrastructure.
The scale of the quarter is worth putting in context. Nvidia’s three-month revenue of $96.2 billion exceeds its full-year revenue in fiscal 2024, which totaled $60.9 billion, and it is roughly double what the company collected in all of fiscal 2023. The company is now adding a business of the size of a large chipmaker every quarter, which is why the financing structures it unveiled matter as much as the products. By guaranteeing leases and power for customers, Nvidia is effectively underwriting the demand that will buy its future chips, a strategy that lets it sell more today while concentrating risk on its own balance sheet.
The concentration of that spending is the risk investors keep weighing. A handful of hyperscalers and AI labs now account for the majority of Nvidia’s revenue, and the company’s own guarantees now tie its fortunes to customers’ ability to pay over multiyear leases. Nvidia executives argued the commitments reflect durable demand rather than speculation, pointing to order books that extend into 2028 and beyond.
The quarter also sharpened the debate over custom silicon. As Nvidia’s revenue swells, its largest customers are designing their own accelerators, and investors have begun asking whether the boom carries the seeds of its own deceleration. The company’s answer, repeated on the call, is that the market is growing fast enough for both approaches to win. The 70 percent target for the coming fiscal year was the clearest version of that argument yet.
China remains the largest unresolved variable. The outlook excludes data center compute revenue from the country, where export controls have limited what Nvidia can sell since 2023 and where the company has relied on specially designed chips for the restricted market. Management said China had been a small part of data center revenue in recent quarters, but the exclusion still complicates a growth target built on the assumption that supply, not demand, is the binding constraint. The company also faces an emerging question closer to home: whether the custom chips designed by its biggest customers will dent its share of inference workloads as models spread beyond data centers into phones, cars, and appliances.
For now, the market is taking the company at its word. The overnight gain put Nvidia’s market value above where it closed before the report, a sign that investors found the growth pledge more persuasive than the margin guide. The real test arrives over the next three quarters, when the Vera Rubin ramp either delivers the fastest product cycle in the company’s history or collides with the power, memory, and financing constraints that come with building AI at this scale.
