NVIDIA reported revenue of $96.2 billion for the quarter ended July 26, 2026, up 106% from a year earlier and 18% sequentially, beating a Bloomberg forecast of $92 billion sec.govaol.com. Data Center revenue was $89.0 billion, up 117% year over year, and the company guided third-quarter revenue to $108.0 billion plus or minus 2% — while assuming no Data Center compute revenue from China sec.gov. GAAP and non-GAAP gross margins were both 75.0%, and GAAP diluted EPS was $2.46 sec.gov. CEO Jensen Huang's summary in the release: "AI has reached its inflection point. It's doing useful work. Its tokens are productive and profitable. Now, compute is revenue" sec.gov.
The quarter extends a streak in which growth is accelerating, not decelerating, as the comparison base rises. Q4 FY26 revenue was $68.1 billion, up 73% year over year; Q1 FY27 was $81.6 billion, up 85%; Q2 FY27 was up 106% sec.govsec.govsec.gov.
| Quarter | Revenue | Y/Y growth |
|---|---|---|
| Q4 FY26 (ended Jan 25, 2026) | $68.1B | +73% |
| Q1 FY27 (ended Apr 26, 2026) | $81.6B | +85% |
| Q2 FY27 (ended Jul 26, 2026) | $96.2B | +106% |
Source: NVIDIA quarterly results releases sec.govsec.govsec.gov.
The platform under the number
The mechanism behind the guide is a generation transition already in volume. NVIDIA said its Vera Rubin platform is ramping into full production, with racks running at CoreWeave, Google Cloud, Microsoft Azure, Oracle Cloud Infrastructure and Nebius, and Spectrum-6 switch systems — supporting both pluggable and co-packaged optics — arriving at what the release calls gigascale AI factories sec.gov. The company also unveiled Vera, which it describes as "the first CPU built for AI agents," and said its Groq 3 LPX interactive inference accelerator is in full production sec.gov. The prior quarter's release framed the succession plainly: Huang said Grace Blackwell with NVLink is "the king of inference today — delivering an order-of-magnitude lower cost per token — and Vera Rubin will extend that leadership even further" sec.gov.
NVIDIA is also extending past silicon. On September 28 it launched the Open Agent Safety Platform, a containment system Huang described as "a browser for agents," with partners including Cisco, Microsoft, Oracle and CoreWeave; VP of enterprise AI Justin Boitano said "model-level safeguards alone can't govern what agents can access or do" cnbc.com. In July it introduced a revenue-sharing financing model for AI clouds, with Sharon AI deploying up to 40,000 GB300 GPUs and Firmus building a Batam campus expected to scale to 360 megawatts and up to 170,000 GPUs barchart.com.
Management is buying its own thesis
On September 29 NVIDIA added $150 billion to its buyback authorization, on top of $80 billion added in May, when the dividend was raised from $0.01 to $0.25 per share cnbc.comsec.gov. The valuation argument behind it: CNBC reported the stock trades at 14.5 times expected fiscal 2028 earnings, below all megacap peers except Micron, against a five-year average P/E of 62.9 — while analysts on average expect roughly $385 billion of net income in fiscal 2028 cnbc.com. Huang told CNBC's Jim Cramer that "buying back Nvidia stock is a tremendous opportunity," and at a Goldman Sachs conference called the company "the world's first and only growth value stock" cnbc.com. The shares closed September 29 at $227.21, with a market cap of $5.476 trillion cnbc.com.
What cuts against it
The bear case is duration and concentration, not demand. Saxo strategist Ruben Dalfovo framed the question the May quarter left open: "The earnings question is no longer whether demand is real. It is whether demand can stay this strong for long enough to justify the expectations already in the share price" home.saxo. The AOL-published report on the August results noted "looming concern about a financial bubble in AI" and that, as big tech spends hundreds of billions on chips and data centers, "the financial benefits remain uncertain" aol.com. China is excluded from guidance entirely, and XTB noted a $4.5 billion H20 inventory charge in FY26 and a potential 25% U.S. import tariff on H200 shipments xtb.com. On the financing side, analysts have flagged circular-financing risk in the revenue-sharing deals — NVIDIA effectively funding demand for its own chips — and GPUs are increasingly traded as a commodity on futures-style contracts barchart.com.
The open question for the November 20 earnings date is whether the third quarter delivers on a guide that already assumes zero China compute revenue — and whether 106% growth, once reported, is finally the quarter that satisfies the multiple sec.govcnbc.com.