Nvidia Guided to $91.0 Billion While Assuming Zero China Data-Center Compute. Wednesday Tests Both Halves of That Sentence.

by Ryo Tanaka · August 25, 2026

Technology Reporter, US Market Current

Nvidia will report results for the second quarter of its fiscal 2027 after the close on Wednesday. The company set the date itself in a July 29 announcement: a conference call on Wednesday, Aug. 26, at 2 p.m. Pacific time, 5 p.m. Eastern, covering a quarter that ended July 26, 2026. Those are company-issued facts, and they are the only part of Wednesday that is not an estimate.

The outlook Nvidia published alongside its prior quarter is equally specific. In the first-quarter fiscal 2027 release, for the quarter ended April 26, 2026, the company said revenue was expected to be $91.0 billion, plus or minus 2%; that GAAP and non-GAAP gross margins were expected to be 74.9% and 75.0% respectively, plus or minus 50 basis points; and that GAAP and non-GAAP operating expenses were expected to be approximately $8.5 billion and $8.3 billion. The company also guided a full-year fiscal 2027 tax rate of 16.0% to 18.0%.

Buried in the middle of that outlook is the sentence that does the most work: "NVIDIA is not assuming any Data Center compute revenue from China in its outlook." Read plainly, that means the $91.0 billion midpoint was constructed with a zero in a line item that was material in earlier periods, and that any approved China data-center compute sales in the quarter would sit outside what was guided rather than inside it.

The base the guidance was built on was large. First-quarter fiscal 2027 revenue was $81.6 billion, up 85% from a year earlier and 20% from the prior quarter, with Data Center revenue of a record $75.2 billion, up 92% from a year earlier and 21% from the prior quarter. GAAP gross margin was 74.9% and non-GAAP gross margin 75.0%. On earnings, the company reported GAAP diluted earnings per share of $2.39 and non-GAAP diluted earnings per share of $1.87 — an unusual ordering, in that the GAAP figure exceeded the adjusted one, and a reminder that the two are not interchangeable when comparing a print against an estimate.

That distinction is the first thing to hold onto Wednesday evening. Consensus figures circulated for Nvidia are generally non-GAAP, and the company reports both. A headline that compares a GAAP result against a non-GAAP estimate, in either direction, is comparing two different things.

The second thing is that consensus itself is not a single number. In a preview published Tuesday, Aug. 25, IG cited a consensus drawn from approximately 60 analysts of $91.9 billion in revenue and $2.09 in earnings per share, and separately referenced an alternative data provider, which IG does not name, showing $93.63 billion and $2.13. IG names LSEG Data & Analytics as a source of analyst projections in the same piece. The figures cited here are as published by IG and accessed on Aug. 25, 2026; compilers assemble different analyst panels on different cutoff dates, and the resulting consensus levels differ accordingly. Anyone treating one of these as "the" number is treating a range as a point.

On the arithmetic as published, the roughly $91.9 billion consensus IG cites sits a little under a percentage point above Nvidia's own $91.0 billion midpoint, and comfortably inside the plus-or-minus 2% band the company drew around it. When guidance and consensus are that close, the revenue line has limited capacity to surprise, and the informational weight shifts to gross margin, to the composition of Data Center revenue, and above all to what the company guides for the following quarter.

IG's preview makes that explicit, arguing that the third-quarter outlook may matter more than the second-quarter result, and noting that management had previously indicated the Vera Rubin platform was in full production with shipments expected to begin in the third quarter. A guidance number that has to absorb a product transition is a different kind of number than one that extends a shipping ramp.

Nvidia's own framing of demand has been unreserved. In the first-quarter release, founder and chief executive Jensen Huang said: "The buildout of AI factories — the largest infrastructure expansion in human history — is accelerating at extraordinary speed. Agentic AI has arrived, doing productive work, generating real value and scaling rapidly across companies and industries." That is a statement about capital spending by Nvidia's customers, and it is the claim the quarter either supports or complicates.

The tape going into the print has been cautious rather than exuberant. Monday's session closed mixed: the Nasdaq composite fell 200.26 points, or 0.8%, to 25,980.19, and the S&P 500 slipped 21.51 points, or 0.3%, to 7,652.86, while the Dow Jones Industrial Average rose 140.15 points, or 0.3%, to 53,417.16 and the Russell 2000 fell 22.79 points, or 0.8%, to 2,995.08, per the Associated Press tally distributed through Barchart. The AP account described stocks drifting to a mixed finish as the countdown ran toward the week's events. Year to date through Monday, the AP figures put the S&P 500 up 11.8%, the Dow up 11.1%, the Nasdaq up 11.8% and the Russell 2000 up 20.7%.

The scheduling is awkward in a way worth naming. Nvidia reports Wednesday after the close, the same day the July PCE report, the second estimate of second-quarter GDP and July durable goods orders are due in the morning, per Newsquawk's weekly calendar; the Jackson Hole symposium then opens Thursday. Two of the largest inputs to how the market prices growth and rates will arrive within about 36 hours of each other, and neither has anything to say about the other.

None of the above is a view on the stock. It is a list of the specific figures the company has already committed to in writing, the specific places the published estimates diverge from each other, and the specific lines in Wednesday's release where the difference between the two will show up first.