Nvidia's quarter: $96.2B revenue, a $108B next-quarter guide, and a record $150B buyback top-up

Nvidia reported fiscal Q2 2027 on August 26, 2026: $96.2 billion revenue (up 106% year over year), 75% gross margin, and $108 billion in Q3 guidance that assumes no data center compute revenue from China. Two days later, the board authorized a $150 billion increase to its share repurchase program — the largest buyback authorization increase in history, per Nvidia. The numbers below come from Nvidia's own press release and filings coverage.

Nvidia's fiscal second quarter of 2027 — ended July 26, 2026, and reported August 26 — is the cleanest signal yet on how the AI infrastructure buildout is pricing itself. Revenue hit $96.2 billion, up 106% from $46.7 billion a year earlier; data center was $89.0 billion, up 117% year over year and now roughly 92% of the total. Gross margin held at 75.0% on both GAAP and non-GAAP bases, and non-GAAP diluted EPS was $2.22 against $2.10 consensus (per LSEG figures in CNBC's live coverage). The guidance set the tone: $108.0 billion for Q3, plus or minus 2%, with Nvidia explicitly not assuming any data center compute revenue from China in the outlook.

What the quarter says

Three numbers carry the story. First, the growth rate is still accelerating on the sequential axis: 18% quarter over quarter after a year of more measured quarter-on-quarter moves. Second, the mix. Data center at roughly 92% of revenue means the consumer and automotive segments are no longer material to the headline; the AI factory buildout is the company. Third, the margin. Holding 75% gross margin while memory prices are soaring industry-wide (CNBC noted Nvidia had "spent heavily ahead to build inventory" for the Vera Rubin ramp) tells you pricing power is still outrunning input cost inflation — for now, on this quarter's costs.

CFO Colette Kress gave analysts a forward number worth marking: she expects fiscal 2028 revenue growth of 70%, against analyst expectations of about 44%, and said customer forecasts "point to our growth doubling next year" while cautioning that guidance reflects supply constraints. CNBC also reported that the stock jumped 4% on the day on the forward guidance. Jensen Huang's framing, from the press release: "AI has reached its inflection point... Now, compute is revenue." That is a demand-side claim — that tokens are "productive and profitable" — and it is what the 70% FY28 expectation rests on.

China: zero assumption, by policy

The guidance line that deserves an underline: Nvidia is not assuming any Data Center compute revenue from China in its outlook. This is not a forecast of zero Chinese sales; it is a deliberate floor. China data center revenue was a meaningful line in earlier quarters, so the guide is a number that stands on the rest of the world alone. For anyone modeling the AI buildout, that makes Nvidia's guide the cleanest available read on ex-China demand — and it also means any policy change on export controls moves the upside, not the baseline.

The $150 billion buyback, in context

Two days after this article's subject quarter, on September 28, 2026, Nvidia's board authorized a $150 billion increase to the share repurchase program, lifting the remaining authorized amount to $235 billion, with execution expected through fiscal year 2028. Nvidia's own release calls it "the largest share repurchase authorization increase in history". The same quarter the company returned about $26.0 billion to shareholders through buybacks and dividends, and it will pay its next $0.25-per-share quarterly dividend on October 1, 2026.

Set that against the balance sheet story in the same filing. Nvidia's Form 10-Q for the quarter ended July 26, 2026, filed August 26, breaks out outstanding indebtedness and the commercial paper program in its liquidity discussion: $25.0 billion of senior unsecured notes issued across seven tranches in June 2026, a $33.4 billion net carrying amount of total debt, and a $25.0 billion commercial paper program with nothing drawn as of quarter end. Maturity profile: $1.0 billion due within a year, $15.0 billion in one to five years, $7.5 billion in five to ten, and $10.0 billion beyond ten years. CNBC's coverage of the same filing reported that the company also provides backstops and other arrangements helping fund new data center buildouts — the mechanism by which an AI chip vendor becomes a capital provider to its own customers. The buyback and the debt are two faces of one strategy: converting record operating cash flow into shareholder returns while financing the buildout that creates it.

What to watch next

• Vera Rubin ramp. The press release puts the Vera Rubin platform into full production, with racks at CoreWeave, Google Cloud, Microsoft Azure, Oracle Cloud Infrastructure and Nebius. Delivery execution here is the physical constraint on the 70% FY28 number.
• Memory costs. A worldwide memory shortage with no sign of abating (per CNBC) feeds directly into gross margin; 75% is this quarter, not a contract.
• Debt trajectory. The 10-Q's liquidity section is now the place to track how much data-center financing Nvidia carries on its own balance sheet versus what partners raise themselves — the notes were only issued in June 2026, so this is a new line item with a short history.
• China assumption. Any loosening of export controls is optionality above a guide that already assumes zero.

Bottom line: Nvidia's fiscal Q2 2027 is the strongest demand document in the AI infrastructure story so far — $96.2 billion, 106% growth, a $108 billion next-quarter floor set without any China compute, and a 70% FY28 expectation from its own CFO. The $150 billion buyback top-up and the $33.4 billion of freshly issued notes are the same story viewed from two directions: extraordinary cash generation being recycled into shareholders and into the financing of the very buildout that creates it.

Prices and benchmark figures are standard published rates as of October 1, 2026 and can change without notice. Some outbound links may be affiliate links — see our disclosure. Found an error? Tell us and we'll correct it in public.