Nvidia Doubled Its Quarterly Revenue. Investors Sold the Margin Guide Instead

Nvidia reported the biggest quarter in its history on Wednesday, and the stock slipped anyway. Revenue hit $96.2 billion for the three months ended July 26 — up 106% from a year ago — with the data center business alone at $89 billion, up 117%. Gross margin held at 75%, meaning Nvidia kept 75 cents of profit on every dollar of sales, a number almost no hardware company has ever produced at this scale. Shares still fell about 1.5% in trading after the report, continuing a pattern that would look absurd if it weren't so consistent: five straight quarters of beating Wall Street's numbers, and the stock was lower the day after four of the five reports before this one.

That pattern is not a market mood. It's the difference between analyzing a quarter and pricing a transition. Spend a few minutes with what Nvidia actually reported, and the selloff stops looking irrational — because the most decision-relevant numbers in the report are the ones that describe what comes next, not what just happened.

The delivered quarter is, by one measure, the best ever

Start with what got delivered, because it sets the bar. Revenue of $96.2 billion was up 18% from the prior quarter and more than double the $46.7 billion Nvidia produced in the same quarter a year earlier. In dollar terms, the company added roughly $15 billion of revenue in a single quarter, following on top of steady growth that has accelerated from about $10 billion of quarterly additions a year ago to $14 billion now — growth in dollars is speeding up even as percentage growth is naturally coming down off a doubling.

The composition matters as much as the total. Data center — AI chips, networking, and the software attached to both — was 92% of everything Nvidia sold. Within that, the fastest-growing slice was no longer just the big cloud companies. Hyperscalers did $49 billion, but the "AI cloud, industrial, and enterprise" bucket did $40 billion, up 138% from a year ago, roughly half the data center total. Nvidia's customer base is widening into neoclouds, sovereigns, and enterprises, not just Microsoft, Amazon, Google, and Meta writing bigger checks.

The company also guided next quarter to $108 billion, plus or minus 2% — another roughly $12 billion step, about 80% above the year-ago quarter and ahead of the roughly $104 billion analysts were modeling. Set that next to the delivered $96.2 billion and there is a simple, defensible reading: one of the greatest growth machines in corporate history is still accelerating.

Why the market shrugged

Now the part that confuses a new investor: how does a company grow revenue 106% and still see its stock sold?

Because at Nvidia's scale, the quarter is no longer the news. Before the report, Nvidia had topped Wall Street's estimates for five straight quarters, and roughly 70% year-over-year growth for the full fiscal year was already baked into the consensus models. At a market capitalization near $5 trillion, flawless execution is the default assumption, not a surprise. Margins of data center revenue at $89 billion came in a hair under the $90 billion level one widely followed analysis had flagged as the threshold for a more constructive view. In other words, the marginal buyer had already decided what a good quarter looked like — and this good quarter met that price without beating it.

This is the "sell the news" mechanics in plain form: when a beat is near-certain, the stock has already moved to where the beat lives. What moves the price is the part of the report that wasn't priced in.

The part that wasn't priced in: a drawn-out margin path

That part came in the outlook. Management didn't just raise the revenue guide to $108 billion. It also guided gross margin to 74.0%, plus or minus half a percentage point, down from the 75% just reported — and then did something unusual: it drew the whole path. Margins are expected to bottom in the fourth quarter in the 71%–72% range, then settle at 72%–73% next fiscal year as price increases the company has already begun executing take effect.

Put a number on what that means. At roughly $100 billion of quarterly revenue, every percentage point of gross margin is worth about $1 billion a quarter. A slide from 75% to a low-70s bottom is roughly $12 billion a year of lost gross profit — meaningful even against the $54 billion of adjusted net income Nvidia produced in this single quarter, and it changes the near-term earnings slope that matters to the next few quarters of stock returns.

Here is the paradox worth understanding. The driver management named is not competition and not a loss of pricing power. It's memory. Nvidia's chips are wrapped in high-bandwidth memory, and the same AI gold rush powering Nvidia's revenue has made that memory scarce and expensive — management cited "extreme pricing conditions" in memory as the primary cause of the step-down. The strongest demand cycle in Nvidia's history has started to raise the cost of its biggest input, which is why the company is reportedly raising prices on some AI server systems by more than 15% even as it ships more than ever. This is cost inflation, not erosion — but it is real, it is quantified by management rather than left as an analyst guess, and it puts a floor on how much of the next leg of growth converts to profit.

This is the cycle turning, and it brings real leverage with it

The margin path matters because it lands at the exact moment the compute cycle Nvidia dominates is changing shape. The Blackwell generation that produced the past year of doubling is handing off to Vera Rubin, the next architecture, which management says began production shipments last month and expects to be roughly 20% of data center revenue in the current quarter. Management calls it the fastest product ramp in the company's history — a claim to be tested by the next two reports, not a result.

The cycle is also moving from training to inference, and that is where Nvidia's competition is real. Management's framing: AI is shifting from prompted chatbots to agents that run continuously and consume far more compute per user. Nvidia is attacking that phase with an inference-specific rack-scale product, the Groq 3 LPX, now in full production, and with Vera, its own CPU for AI servers. But inference is exactly where the economics change — the game shifts from raw training domination to latency, cost per token, and efficiency, which is where custom chips from hyperscalers and AI labs are aimed. The question was never whether Nvidia wins the training phase. It already did, to the point of a near-monopoly. The question is at what speed and at what margin it defends the inference phase.

Underneath the growth sits a series of commitments that cut both ways. Nvidia holds roughly $119 billion in supply-related commitments, and it is helping raise $500 billion of third-party capital through partnerships with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR. Management says the top five hyperscalers plan roughly $800 billion of capital spending this calendar year and $1.3 trillion next, and that its own outlook for about 70% growth next fiscal year is "supply-constrained" rather than demand-limited. Demand is emphatically real. But every one of those arrangements is a promise that concentrates pain if any major buyer's budget pauses — supply commitments are a dual signal. They read as durability while the boom runs, and as leverage the moment it stalls.

What this report actually decides

Take stock of where the evidence lands. The Q2 report settles the demand question as emphatically as it has ever been settled: revenue doubling, a data center business at $89 billion a quarter, the company's own CFO describing compute as fully utilized across every cloud Nvidia serves. The report does not settle the next two quarters, where management has disclosed a margin slide of three-plus points before a partial recovery, and it does not settle the Rubin transition, which is now a roadmap claim awaiting delivery.

Those are the two numbers to watch, and they are observable: does gross margin actually bottom in the 71%–72% range and lift as price increases flow through, and does Vera Rubin go from "fastest ramp in our history" to a reported circa-20% of data center revenue, and then more? That is the milestone that separates a claim from a financial result.

For an investor deciding whether the stock belongs in a portfolio, the frame is cleaner than the noise suggests. The long-term thesis — Nvidia still sits at the center of the compute cycle it created — remains intact, and a selloff on a strong report has historically been a buying opportunity rather than a verdict. But at a $5 trillion market value, roughly 40 times trailing earnings and about 23 times this quarter's annualized adjusted profit, the market is already paying full price for flawless delivery. The margin guide says the near-term earnings slope is a little lower than a straight-line read of 75% would imply, and the transition carries execution risk that no amount of revenue momentum removes. The judgment is not whether AI demand is real — that question is closed. It is whether the return over the next several quarters, through a known margin dip and an unproven ramp, is still the best use of capital that has other homes in this cycle.