Deep Dive — Nvidia Q3 results: $115.8B revenue, $130B guide

Share
Deep Dive — Nvidia Q3 results: $115.8B revenue, $130B guide

Nvidia's first quarter without Jensen Huang was its best ever. Revenue for the three months to October 25 came in at $115.8 billion, up 103% from a year earlier and $6.7 billion above the $109.1 billion analysts had modelled, according to the company's results release. Data Center revenue was $107.6 billion, up 110%. Adjusted earnings per share of $2.68 beat the $2.48 consensus by 20 cents, CNBC reported, and the stock rose more than 3% after hours.

Then the company guided the January quarter to $130 billion, told investors gross margin would fall to 71–72%, and said it assumes no Data Center compute revenue from China at all.

The beat answered the question the market spent two weeks asking: does the machine run without its founder? It does. The guidance answered a question almost nobody asked out loud: what does the next leg of growth cost, and who is paying for it?

The quarter itself was clean

The composition is what you would expect from a company selling into a shortage. Data Center was 93% of revenue. Hyperscale customers — the big clouds — bought $60.1 billion of it, up 71% from a year earlier. The segment Nvidia labels AI Clouds, Industrial and Enterprise bought $47.5 billion, up 195%, Zhidx reported from the call. That second number is the one to keep in mind later: the customers growing fastest are the ones with the thinnest balance sheets.

Gross margin landed at 74.9% on a non-GAAP basis, above the 74% analysts expected but below the prior quarter, which Nvidia attributed to the cost of ramping Vera Rubin, per SiliconANGLE's account of the call. Rubin added nearly $20 billion of revenue sequentially as it ramped — the fastest product ramp in company history, Kress said, as Zhidx reported it. GAAP net income was $65.3 billion, up 105%. The board declared a quarterly dividend of $0.25 per share, payable December 23.

For a company that had just lost the chief executive who set its allocation policy, its pricing and its partner relationships, none of that is normal. It is the balance sheet of a business with a two-year order book and no intention of slowing down for a personnel problem.

The guide is where the argument starts

Two details in the January-quarter outlook do most of the work.

The first is the calendar. That quarter has 14 weeks, not 13 — the release says so in the outlook section. Divide the $130 billion guide by 14 and you get $9.29 billion of revenue per week. Divide this quarter's $115.8 billion by 13 and you get $8.91 billion. Sequential growth per week: 4.2%. The same arithmetic on the quarter just reported gives roughly 20%.

That is a real deceleration, and it is not a forecasting error on Nvidia's part. It is the shape of a company that has sold out its capacity and is now guiding on what it can physically ship. The extra week also flatters the year-over-year number CNBC cites — roughly 90% growth in the January quarter — since the 14-week period has no 14-week base to beat.

The second detail is margin. Nvidia expects 71–72% gross margin in the January quarter, down from 74.8% GAAP this quarter, and says that quarter is the trough before a recovery to 72–73% across fiscal 2028. On $130 billion of revenue, three points of margin is about $4 billion of gross profit given up relative to holding this quarter's rate. Revenue grows 12% sequentially; gross profit grows about 7%.

Kress put the reasoning in the guidance itself, saying it "considers data center readiness, the strength and durability of customer demand and our customers' access to capital," as SiliconANGLE reported. Read that list again. Two of the three inputs are not demand signals. They are logistics and credit.

China went to zero in one step

Buried in the call was the cleanest illustration of how fast the policy environment is moving under this company.

Nvidia booked $2.4 billion of licensed Hopper data-center sales into China in the quarter just reported, and is no longer shipping those products there, Kress said, per Zhidx. The January guide contains none of it. That is a product line going from a couple of billion a quarter to zero inside one reporting period, which the company absorbed by guiding revenue up anyway.

The rule that closes the rental route landed the day before the call: Commerce published an interim final rule that makes renting American compute from a third-country data center a licensable act when the customer is Chinese, as we covered when it published. CNBC reported that the change would prevent Nvidia chips from being rented to Chinese companies entirely. Nvidia's response was to raise its fiscal-2028 growth forecast from 70% to 75% — the number that had been the single biggest open question on the call — and to say the outlook does not depend on China at all.

That is either confidence or a company that has decided the China question is now a legal matter rather than a commercial one. Both readings are available, and the market took the first: a stock down 24% this month, coming off six straight weekly declines, its longest losing streak since 2012 and its lowest level since July 2025, added 3% after the print.

Who is actually paying for the growth

This is the part of the story the earnings coverage mostly skipped, and it is the part that decides whether the 75% forecast is a promise or a hope. Our preview before the print argued the revenue estimate was the least interesting thing Kress would say, and that the guarantee language was where the risk sat. That held.

Start with who is buying. Frontier labs — OpenAI, Anthropic and their peers — will account for about 15% of Data Center revenue this year, Kress said. Anthropic's annualized revenue reached $181 billion at the end of October, she added, per SiliconANGLE. Those numbers are the demand side of the AI economy stated in one breath: a lab that did not exist as a serious enterprise vendor three years ago is now running at a revenue rate larger than most Fortune 100 companies, and it is renting most of its compute.

Now the financing side. Nvidia has more than $32 billion of commitments to neoclouds through 2032, Kress said, and those commitments only take effect if the customers secure the funding to buy the hardware and the compute capacity is deployed and operating. Zhidx put the wider stack of obligations at $390 billion of purchase commitments, more than $100 billion of data-center leases and cloud commitments, and $118 billion of guarantees — including $105 billion of credit support for SB Energy's Ohio site, whose 20-year lease supplies OpenAI.

The scale of that last number is the story. Nvidia is not merely selling to companies building AI infrastructure; in the largest case it is standing behind the credit that lets one of them build at all. Management's defense is that this is conditional, that guarantees activate only after financing closes and capacity is operational, and that the disclosure language was clarified rather than changed. That is a fair reading of the mechanics. It is also a description of a company whose revenue depends, at the margin, on its own willingness to underwrite its customers.

The cash-flow statement shows the same thing in a less flattering form. Nvidia spent $15.7 billion buying equity securities in the quarter, against $3.7 billion a year earlier, and $58.1 billion in the first nine months of the fiscal year. Accounts receivable absorbed $18.05 billion of cash, against $5.58 billion a year earlier — a rise almost exactly the size of the $19.3 billion sequential revenue increase. Cash and marketable securities fell to $53.5 billion from $56.6 billion in July.

None of that is accounting misconduct. Vendor financing has been a legitimate tool since the mainframe era, and it has also been the mechanism that broke hardware companies when the cycle turned — Lucent's customer loans in 2000 are the standard reference for a reason. The relevant fact here is narrower and harder: Nvidia added $150 billion to its buyback authorization, leaving $227 billion to execute by the end of fiscal 2028, while its cash balance fell and its receivables grew almost in step with revenue. A company buying back that much stock while its working capital expands that fast is telling you it expects the cash to keep arriving.

The governance question the numbers cannot answer

Harvey Jones, the board's first-ever chair, opened the call by saying Huang resigned because his conduct fell short of the company's code and that the board "does not intend to reinstate Jensen as CEO," per Zhidx's account. A search has started for a permanent chief executive, considering internal and external candidates including Kress, and the board added a national security and export control committee. Huang had posted hours earlier that he would be honored to return if the board judged it in the shareholders' and the country's interest.

The awkwardness was visible. Kress, five days into the job, misstated a bank's name and ended the call ten minutes early; Ian Buck, the CUDA creator the board put on the call, did not speak. "Business as usual at Nvidia," she said, and on the numbers she was right.

But the quarter that produced this beat ended October 25 — before the resignation, before the Commerce rule, before the Senate briefing Huang was asked to attend. The January guide is the first set of numbers that prices the new political reality in. It is also the first set produced by a chief executive who will not be running the company when the quarter closes, which is why the reinstatement question is still open no matter what the chair says on a call.

On top of that sits the Securities and Exchange Commission's civil investigation into whether Nvidia's disclosures misled investors, first reported by the Wall Street Journal and summarized by Reuters. An investigation into disclosure, running in the same quarter that the company changed the wording around $118 billion of guarantees, is not a rounding error.

The case against the bear case

The honest counterargument deserves space, because the skeptical read has a hole in it.

If Nvidia were financing demand it could not otherwise find, you would expect the demand to be soft at the edges: customers delaying, inventory building, price cutting. What the quarter shows is the opposite. Supply is the constraint — TSMC executives visited Nvidia's offices this week, Kress said, and the company has locked memory supply and advanced packaging capacity, per Zhidx. Purchase commitments of $390 billion are not a demand problem; they are a company betting its balance sheet that it can sell everything it can make.

And the margin guide is not a warning about pricing power. It is the cost of a node transition: 74.9% to 71–72% during a Rubin ramp, with management guiding the rate back to 72–73% through fiscal 2028. Nvidia still earns software-like margins on hardware at a scale no semiconductor company has reached. The bears who predicted margin collapse at every generation are still waiting.

The defensible concern is narrower: growth at this rate requires customers who can pay, and the customers growing fastest — neoclouds, labs — are the ones whose ability to pay depends on capital markets staying open and on Nvidia's own participation in those markets. That is a cyclical exposure dressed as a structural one, and the guidance's own language about "customers' access to capital" concedes it.

What to watch

A named chief executive. The search is the single cleanest signal on whether this is a transition or an interregnum. An internal appointment — Kress or Buck — reads as continuity; an outside hire reads as a board that wants a different relationship with the founder.

The January margin. 71–72% is now a promise. If the quarter lands at the bottom of that range with Rubin volumes climbing, the "trough" framing starts to look optimistic, and the fiscal-2028 recovery to 72–73% becomes the next number to defend.

The SEC probe's scope. A disclosure investigation that stays narrow is a legal cost. One that widens into the guarantee structures is a different company.

The Commerce rule's effective date, and December 10. The rental rule closes the channel that got Huang fired; Commerce Secretary Howard Lutnick testifies before Senate Banking on December 10 about whether his department is enforcing the controls. If a rule with a date attached lands, the case for reinstating the architect of the old channel loses its commercial logic, whatever the petition says.

The $390 billion of purchase commitments. This is the number that turns a demand story into a supply story. If Nvidia is still adding to it in January while revenue per week grows 4%, someone is building inventory ahead of orders.

The quarter proved the machine runs without its founder. The guide proved what running it costs: three points of margin, a working-capital build, and a growing willingness to stand behind customers' credit. Whoever takes the permanent job inherits a company that can still grow 90% a year — and a set of obligations that only work if everyone else's financing does too.

If you were on Nvidia's board, would you hire a CEO from inside the company or outside it? Write to us via our contact page.

Get SiliconNoon in your inbox

One email a day with the AI stories that matter — what happened, the numbers, and why it matters. No spam, unsubscribe in one click.

Read more

Jensen Huang resigns: Colette Kress is Nvidia's interim CEO

Jensen Huang resigns: Colette Kress is Nvidia's interim CEO

Nvidia's board announced Thursday that Jensen Huang has resigned as president, chief executive and a member of the board, effective immediately, the company said in a statement. Colette Kress, chief financial officer since 2013, becomes interim CEO and keeps the CFO job. Harvey C. Jones, the former Synopsys chief executive, was appointed the first chair of the board in Nvidia's 33-year history. The statement carries two quotes: one from Jones, one from Kress. Neither is from Huang.

Deep Dive — Anthropic's IPO opened at $111.50 and closed at $97.17

Deep Dive — Anthropic's IPO opened at $111.50 and closed at $97.17

Three prices explain Anthropic's first day as a public company, and only one of them is the number the company chose. The offering priced at $83 on Wednesday night. The first trade printed at $111.50 on Thursday morning, 34.3 percent higher. The close was $97.17, up 17.1 percent, and by then the stock had surrendered half of everything it gained at the bell — $14.33 of a $28.50 pop. That $28.50 is the standard measure of what the issuer left behind.

Deep Dive — Baidu Token Exchange: verified models, uninsured buyers

Deep Dive — Baidu Token Exchange: verified models, uninsured buyers

Baidu spent the past month quietly building a marketplace for other companies' models, then told the world about it this week. Token Exchange — 词元所, Ciyuansuo — went live at Baidu World 2026 on November 12, roughly a month after it began taking suppliers. Robin Li introduced it as "an exchange for large language models," and the analogy he reached for was a stock market: a stock exchange attracts good companies to list so investors can find and back them; Token Exchange, he said, gathers good…

Fractile raises $720M at $7B, six times its May valuation

Fractile raises $720M at $7B, six times its May valuation

Fractile, the Bristol- and London-based chip designer, has closed a $720 million Series C at a $7 billion post-money valuation, according to its investor Balderton. That is roughly six times the $1.1 billion it was worth in May, according to the Financial Times, when it raised a $220 million Series B — and it makes Fractile one of the most valuable technology companies in Britain, with a first commercial chip still a year from data centres. The round was co-led by Thrive Capital, Founders…

Deep Dive — OpenAI's $30B round fell to $20B at a $1.25T valuation

Deep Dive — OpenAI's $30B round fell to $20B at a $1.25T valuation

In September, OpenAI went to investors asking for $30 billion at a $1.4 trillion valuation. On Thursday night the Financial Times reported the target is now about $20 billion at $1.25 trillion, because some investors cut the size of their commitments amid the AI selloff, according to the FT. Hours earlier, the Wall Street Journal had put a number on how far the round has actually got: roughly $11 billion committed, less than half the original target, the Journal reported. That is the single…

OpenAI plans to buy Google TPUs as Oracle starts selling them

OpenAI plans to buy Google TPUs as Oracle starts selling them

For ten years the TPU was the one advantage Google refused to sell outside its own walls. This week it sold the distribution, and within a day the buyer on the other end was Nvidia's largest customer. Oracle will offer Google's tensor processing units through Oracle Cloud Infrastructure starting in calendar 2027, the first time TPUs will run on a hyperscale cloud other than Google's own, Oracle announced at its AI World conference.

Deep Dive — AI neoclouds hit a credit wall as SpaceX borrows $40B

Deep Dive — AI neoclouds hit a credit wall as SpaceX borrows $40B

On Sunday evening, Elon Musk announced on X that SpaceX had closed $40 billion of debt to buy Nvidia GPUs, and used the announcement to deliver a verdict on his competitors: the debt markets have "almost entirely shut off access to money for any company that doesn't have substantial revenue independent of AI data centers," and "the 200+ neoclouds are going to see guppy-level death and consolidation," he wrote. The $40 billion is the number everyone will quote.