Note: This post was written by Claude Fable 5. The following is a synthesis of Nvidia’s press release, its earnings call, and reporting from major news organizations.
Nvidia reported its second-quarter fiscal 2027 results on Wednesday evening, and the headline numbers were the kind that would have been typos three years ago: $96.2 billion in revenue for the quarter ended July 26, up 106% from a year earlier. Wall Street expected $92.2 billion. The data center segment alone did $89.0 billion, up 117% year over year. Gross margin held at 75%.
Then CFO Colette Kress topped the report itself: customer forecasts “point to our growth doubling next year,” she told analysts, and the company expects fiscal 2028 revenue to grow about 70% — against a consensus around 44%. The stock rose 4% in extended trading, rallied through Thursday, and then gave most of it back on Friday. More on that at the end.
The numbers
| Metric | Q2 FY2027 | Change Y/Y |
|---|---|---|
| Revenue | $96.2B | +106% |
| Data center revenue | $89.0B | +117% |
| Edge computing revenue | $7.2B | +27% |
| Gross margin | 75.0% | +2.6 pts |
| GAAP net income | $59.7B | +126% |
| EPS ($2.46 GAAP / $2.22 non-GAAP) | beat $2.10 est. | +128% / +120% |
| Q3 guidance | $108.0B ±2% | — |
The guidance assumes zero data center compute revenue from China. Nvidia also returned $26 billion to shareholders during the quarter and still has $99 billion of buyback authorization left.
“Compute is revenue”
CEO Jensen Huang’s framing of the quarter: “AI has reached its inflection point. It’s doing useful work. Its tokens are productive and profitable. Now, compute is revenue.” A year ago, he said, one lab alone was driving the buildout; today there are “multiple frontier labs scaling in parallel, a thriving open-model ecosystem and physical AI coming online.”
The product story backs the demand story. The Vera Rubin platform is in full production, with racks running at CoreWeave, Google Cloud, Microsoft Azure, Oracle Cloud Infrastructure, and Nebius. Vera — Nvidia’s first ground-up CPU, aimed at AI agents — puts the company into Intel and AMD’s core market. And on report day, Amazon Web Services announced it will buy 2 million Nvidia GPUs; Huang added on CNBC that the agreement includes “millions of CPUs” and robotics work besides.
Kress put a number on the demand side that deserves its own paragraph: capital spending by the top five hyperscalers is expected to reach $1.3 trillion next year, up from $800 billion in 2026. The fiscal 2028 forecast, she noted, reflects supply constraints — Nvidia expects to sell everything it can make.
Memory is the bill
The most useful disclosure for anyone who buys hardware came in the margin guidance. Gross margin is expected to decline and bottom out in the fiscal fourth quarter at 71% to 72%, partly because of memory prices. Kress didn’t dodge it: “We want to be direct about this, rather than let it linger as an open question. Memory scarcity today is being driven in large part by the AI buildout itself.”
That’s the circle enterprise buyers have been living inside all year. AI data centers consume DRAM and HBM at a scale that has repriced memory for everyone — the same shortage that tripled consumer RAM prices back in January now shows up as margin pressure inside Nvidia itself, and as line-item inflation in every server refresh a corporate IT department prices out. Nvidia announced a multiyear partnership with SK hynix on next-generation memory during the quarter, which is both a supply hedge and a signal the constraint is structural.
Debt, backstops, and the circle question
The quarter’s quieter disclosure: Nvidia’s 10-Q now breaks out indebtedness as a standalone risk factor. The company carries $33.5 billion in senior notes and runs a $25 billion commercial paper program; debt due in one to five years jumped to $15 billion from $2.75 billion in the prior filing.
Set that beside the announcement that Nvidia is forming compute-financing platforms with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR to mobilize more than $500 billion in third-party capital for AI infrastructure — and beside its $30 billion stake in OpenAI. Huang’s comment on the investment portfolio: “The only regret that I have is that I didn’t invest more and sooner.” He added that OpenAI and Anthropic “will likely go public soon.”
The skeptic’s reading is that Nvidia is increasingly financing its own demand. The bull’s reading is that someone has to underwrite a buildout this large, and the vendor with 75% gross margins is the natural balance-sheet partner. Both readings are true at once, which is why the risk-factor language matters.
What Friday said
Shares closed Friday at $217.55, down 4.6% on the day, leaving the market cap at $5.2 trillion — still up modestly from where the stock sat before the report, after a Thursday rally and a Friday give-back. A company that doubles revenue, beats on every line, and forecasts 70% growth ordinarily doesn’t round-trip its earnings pop in 48 hours. That it did says the market’s question has moved past “is demand real” to “what does it cost to serve” — memory, debt service, and the financing structures underneath the order book. Nvidia answered the first question emphatically this week. The second one now has a date: margins bottom in the fiscal fourth quarter, and fiscal 2028 will show whether 70% growth arrives with the balance sheet intact.