NextFin News - Nvidia reported second-quarter revenue of $96.2 billion, up 106% from a year earlier and well above the roughly $92 billion Wall Street expected, then guided the current quarter to $108 billion and, for the first time ever, forecast a full year out with 70% growth. The report pushed shares about 7% higher in premarket trading Thursday, lifted Nasdaq 100 futures 1% and S&P 500 futures 0.4%, and did what three months of chip-stock selling had failed to do: it calmed the market's anxiety that the AI buildout was running out of buyers.
The anxiety had a name - peak AI. For a quarter, investors fretted that the hundreds of billions of dollars flowing into data centers would not generate returns, and that hyperscaler capital spending was about to roll over. Nvidia's print, and especially its unprecedented year-ahead guidance, argued the opposite. Demand is not just intact; it is accelerating faster than supply can respond. The chipmaker's own guidance is now "supply-constrained," in the words of Chief Financial Officer Colette Kress: customer forecasts point to demand potentially doubling, but Nvidia is holding its own forecast to 70% because it cannot build chips fast enough.
The Numbers That Moved the Market
For the quarter ended July 26, 2026, Nvidia said revenue reached $96.2 billion, up 18% from the prior quarter's $81.6 billion and more than double the $46.7 billion posted a year earlier. GAAP and non-GAAP gross margins both landed at 75.0%. GAAP earnings per diluted share were $2.46; non-GAAP earnings per diluted share were $2.22. GAAP net income rose 126% year over year to $59.7 billion, and operating income climbed 124% to $63.7 billion.
Data Center revenue - the segment that carries the AI trade - generated $89.0 billion, up 117% from a year ago and the overwhelming majority of the total. Within it, accelerated computing and networking revenue reached $40 billion, up 25% sequentially and 138% year over year. Networking revenue grew 18% sequentially, with Spectrum-X Ethernet up 2.6 times from a year earlier. Sovereign AI revenue grew 35% sequentially and more than tripled year over year, a sign that the buildout is broadening beyond the U.S. hyperscalers that dominated the first wave.
The guidance mattered as much as the print. Nvidia expects third-quarter revenue of $108 billion, above the roughly $104 billion consensus, which would be the company's first quarter above $100 billion in revenue. And in a break from its own practice - "we've never forecasted, never guided to a year in advance," CEO Jensen Huang said on the earnings call - management said it expects revenue to grow approximately 70% in fiscal 2028. Applied to the current-year run rate, that implies fiscal 2028 revenue in the range of $690 billion to $700 billion, more than $100 billion above the roughly $570 billion analysts had been modeling.
There was one cautionary note embedded in the release. For the third quarter, Nvidia expects GAAP and non-GAAP gross margins of 74%, plus or minus 50 basis points - a modest step down from the 75.0% printed in the second quarter, reflecting the mix shift as new architectures ramp and memory costs rise. Some analysts have modeled a further reset to roughly 71% by the fourth quarter.
"AI has reached its inflection point. It's doing useful work. Its tokens are productive and profitable. Now, compute is revenue," Huang said in prepared remarks. "And demand is accelerating. This time last year, one lab alone was driving the buildout; today, we have a golden age of new AI labs and startups, multiple frontier labs scaling in parallel, a thriving open-model ecosystem and physical AI coming online - with strong momentum across the U.S. and around the world. The AI infrastructure buildout is at full steam."
The company also returned approximately $26.0 billion to shareholders in the quarter through repurchases and dividends, leaving about $99.0 billion under its buyback authorization. Nvidia will pay its next quarterly cash dividend of $0.25 per share on October 1, 2026.
Why One Earnings Print Reversed a Three-Month Selloff
The question worth asking is why a single company's results could reverse a broad, three-month decline across AI-linked equities. The answer runs through the capital-expenditure feedback loop that has driven the entire AI trade since 2023 - and through the risk premium, not the income statement.
Hyperscalers - Microsoft, Amazon, Google, Meta - justify their AI infrastructure budgets by pointing to revenue growth in their cloud businesses. If cloud growth slows, the logic for continued spending weakens, and the entire supply chain gets re-rated lower: not just chips, but memory, networking, power, cooling, and the miners that monetize compute capacity. Nvidia's revenue is, in effect, a real-time meter of how much money is still flowing into AI infrastructure. A 106% year-over-year increase, with margins holding at 75%, is evidence that the hyperscalers have not pulled back. Their capex is still arriving as orders.
That is the first-order effect. The second-order effect is what actually moved the market. Semiconductor, memory, and AI infrastructure stocks had been sold down not because their earnings had deteriorated - most were still printing strong results - but because investors feared earnings were about to deteriorate. That fear lives in the discount rate applied to future cash flows, not in the current quarter's numbers. When the feared deterioration did not appear, the risk premium compressed, and valuations re-rated higher across the group even before the next round of fundamentals arrived. The iShares Semiconductor ETF rose 3% before the bell; the Roundhill Memory ETF advanced 3.5%; memory and networking names followed; Salesforce and CrowdStrike, also reporting, jumped 11% and 9%.
The transmission channel, in other words, was not better earnings this quarter. It was a lower discount rate applied to earnings two years out. That distinction matters, because it means the rally is more fragile than it looks: it rests on expectations being sustained, not on a single print.
There was also a cross-asset channel. Bitcoin topped the $80,000 threshold for the second time in a week, and AI infrastructure miners advanced - IREN rose 5%, while TeraWulf and Cipher Mining each gained around 3%. The linkage is not mystical. These companies are, in effect, leveraged calls on the price and availability of compute capacity. When the market's confidence in the AI buildout strengthens, their revenue visibility improves, and their shares re-rate with the broader complex. Bitcoin, increasingly framed as a liquidity-sensitive risk asset, tends to move with the same risk-on flows.
Cyclical Wave or Structural Shift - The Call That Determines the Trade
The central analytical question is whether the AI infrastructure buildout is a cyclical investment wave that will mean-revert, or a structural regime shift that will not correct on its own. Getting this wrong flips the conclusion, so it is worth separating the two forces rather than blending them.
The evidence for a structural shift is substantial. Nvidia's management is projecting 70% revenue growth two fiscal years out - a horizon the company has never before guided to. Customer forecasts point to demand doubling. The company is working with BlackRock, Blackstone, KKR, Apollo, Brookfield, and Goldman Sachs to establish a $500 billion pool of third-party capital to finance AI compute infrastructure, a move designed to securitize GPUs and turn them into a financeable asset class. Nvidia is also backing SB Energy and OpenAI's effort to build an 8-gigawatt data center in Ohio, with up to $150 billion in investment. These are not the actions of a company managing a cyclical peak; they are the actions of a company building the plumbing for a new computing paradigm.
The evidence for a cyclical overlay is equally real. The buildout is ultimately funded by hyperscaler capex budgets, which are discretionary and historically cyclical. Each major cloud provider is developing its own AI silicon to reduce dependence on Nvidia - a structural headwind on Nvidia's share of the pie even if the total pie keeps growing. And customer concentration is rising: two customers combined accounted for 39% of Nvidia's total revenue in the quarter, with the largest alone contributing 23%, up from 16% and 14% in the fiscal first quarter, according to the company's investor disclosures. Concentration of that kind is a fragility, not a strength - without those two accounts, revenue would have declined quarter over quarter.
The judgment: the demand driver is structural - the transition to AI computing is a multi-year regime shift in how computing workloads are architected - but the pace of infrastructure spending is cyclical and will oscillate with hyperscaler budget cycles, interest rates, and the realized return on AI investments. This is a structural trend with a cyclical amplitude. Investors who treat every quarterly wobble as the end of the AI boom will miss the trend; investors who treat the trend as a straight line will be caught by the oscillations.
What the Market Has Already Priced In
A second-order question the market is not asking loudly enough: how much of the structural case is already in the stock? Nvidia is the world's most valuable publicly traded company, and its valuation embeds years of continued hypergrowth. The 70% fiscal 2028 forecast landed above the roughly $570 billion analysts had modeled, which is why the shares rallied. But a forecast that beats consensus today becomes the baseline tomorrow.
This is the trap that has caught Nvidia investors before. The company has beaten estimates in five of the last six quarters, yet the stock fell after four of those beats. The pattern is not a mystery: when expectations are set this high, "good" is not enough - the news has to be good enough to justify a higher multiple, and that bar rises every quarter. A stock can fall on good news if the news is not good enough.
The valuation question cuts both ways. Bulls argue that a company delivering 70% growth at 75% gross margins deserves a premium multiple, and that the $500 billion financing platform could unlock a new pool of institutional capital that the market has not yet underwritten. Bears argue that the multiple assumes the capex cycle never pauses, and that any pause - even a temporary one - would compress the multiple even if earnings keep rising. Both can be right at different points in the cycle.
The Strongest Case Against the Bullish Read
The bear case is not a fringe view, and it should not be dismissed as such. It is the position that drove the three-month selloff in chip stocks, and it rests on a sound observation: Nvidia's results are backward-looking, and the stock prices in forward growth. Capital spending of this magnitude must eventually produce cash flows, and the cash flows visible today remain small relative to the investment. If hyperscaler capex growth slows even modestly - if the marginal dollar of AI spending generates less revenue than expected - the multiple can compress even while earnings keep rising.
There is also the margin question. Nvidia's own guidance calls for 74% gross margins in the third quarter, down 100 basis points from the second quarter, and some analysts have modeled a further reset to roughly 71% by the fourth quarter as new architectures ramp and memory costs rise. Margin compression of that kind would not invalidate the growth story, but it would compress the earnings power that justifies the valuation.
And there is the China question. Nvidia's third-quarter guidance of $108 billion assumes zero Data Center compute revenue from China, following export restrictions on advanced AI chips. The company has benefited from the release of restricted inventory to customers outside China in prior periods, but that is a one-time tailwind, not a recurring revenue stream. A guidance number that excludes an entire geography is only as strong as the demand from the geographies that remain.
The falsifying signal is specific and observable: watch Nvidia's data center compute revenue quarter over quarter, and watch whether the company can sustain 70%+ growth while guiding with zero China compute. If compute revenue flattens or declines for two consecutive quarters while guidance keeps rising, the structural-demand thesis is wrong. It would mean growth is being carried by one-time items - inventory releases, networking, sovereign AI - rather than by durable, recurring compute demand. That threshold, not the headline revenue number, is what breaks the bullish case.
What to Watch, and Who Is Exposed
The beneficiaries of a sustained AI buildout extend well beyond Nvidia: semiconductor equipment makers, memory producers, networking vendors, power and cooling providers, and the AI infrastructure companies that monetize compute capacity. The exposed are the companies that have priced in uninterrupted growth - any name whose valuation assumes the capex cycle never pauses - and the hyperscalers themselves if their AI spending fails to generate the returns they are underwriting.
By time horizon, the watchlist splits cleanly:
- Short term (sentiment and liquidity): follow-through earnings from the chip supply chain. A string of strong prints would confirm that the risk-premium compression is broad-based, not Nvidia-specific. Marvell Technology reports after the bell Thursday and will offer the next read on AI networking demand.
- Medium term (fundamentals): hyperscaler capex guidance in the coming quarters, and whether cloud revenue growth keeps pace with spending. This is the actual constraint on the cycle. If cloud growth trails capex growth for two consecutive quarters, the feedback loop reverses.
- Long term (structural): whether the $500 billion financing platform and similar vehicles successfully securitize AI compute, turning GPUs into an institutional asset class. If they do, the buildout can outlast any single hyperscaler's budget cycle.
Three scenarios frame the path:
- Base case: AI capex continues to grow through fiscal 2028 at a decelerating but still-high rate; Nvidia delivers on its 70% growth projection; the market rewards the supply chain selectively rather than uniformly.
- Upside case: customer demand doubles as forecast, supply constraints ease faster than expected, and the financing platform unlocks a new pool of institutional capital. AI infrastructure names re-rate materially higher.
- Downside case: hyperscaler capex growth slows as AI revenue disappoints, compute revenue flattens for two consecutive quarters, and the risk premium on the AI complex widens again. The multiple compresses even as earnings rise.
In a separate reported development, Nvidia has agreed to acquire AI developer platform Hugging Face, according to a person familiar with the matter, in a deal that would broaden the chipmaker's reach into the software and open-source model ecosystem. Neither company had commented at the time of reporting, and the transaction remains subject to the usual uncertainties of an unannounced deal.
Nvidia did not just beat estimates; it bought the AI trade time. Whether that time converts into returns depends less on the next earnings print than on the cash flows the spending eventually produces.
Market data as of premarket trading Thursday, August 27, 2026. Nvidia shares closed Wednesday at $209.76 before the report.
Explore more exclusive insights at nextfin.ai.
