Nvidia has beaten consensus in every quarter of this AI cycle. Its shares have then fallen the day after each of the last four prints. That pattern, rather than the revenue line, is the thing to watch on Wednesday.
The basics
Results land after the US close on 26 August, with the call at 10pm UK time. Nvidia guided second-quarter revenue to $91bn, plus or minus 2%. Consensus has crept up to about $91.9bn, with adjusted earnings near $2.08 a share. Both figures are about double the same quarter last year. Gross margin is guided at close to 75%. The starting point is a first quarter that delivered $81.6bn of revenue, of which the data centre division supplied $75.2bn. Growth of 95% year on year no longer settles anything.
Third-quarter guidance is the real test
Traders at moomoo reckon a figure near $105bn is needed to satisfy the room. Anything starting with a nine invites awkward questions about whether the Vera Rubin handover creates an air pocket. Rubin entered full production earlier this year, with shipments due in the second half.
China stays excluded from the guide
Nvidia assumed zero China data centre compute revenue, so any commentary is pure upside. ByteDance and Tencent have each taken about 10,000 H200 chips, the Financial Times reported, against a per-customer ceiling of 75,000. Licences exist. Chips are barely moving. Beijing, not Washington, now holds the pen.
Memory costs are pressing on margins
Server DRAM contract prices rose 53% to 58% in the second quarter, according to TrendForce, with more to come. Bloomberg reported that Nvidia has warned customers of AI server price rises above 15% next year. A company with 75% margins passing costs through tells you who holds the whip hand in memory.
Rates, oil and the circular financing question
The 10-year Treasury yield sits near 4.7%, Brent is above $90, and the July PCE reading lands the same day. Nvidia is 7.3% of the S&P 500, so this is an index event as much as a company one. A securities filing showed guarantees of up to $105bn on lease obligations tied to an OpenAI campus in Ohio. Critics keep reaching for the word circular.
Options are pricing a move of 5% to 6% in either direction. Nvidia has slipped the day after each of its last four reports, by an average of about 2.8%, and the shares have shed 7.5% across the current streak. Analysts have not blinked. Fifty-nine of 62 rate the stock a buy, and the average target sits far above a share price hovering near $210 and a market value of $5.2trn.
Wednesday does not hinge on whether Nvidia beat. It hinges on whether Jensen Huang can describe demand that Nvidia is not, in some part, financing itself.
Be careful. Here's our fact-check notes
- Consensus figures conflict across sources. Revenue estimates range from $91.85bn (Rex Shares, 40 analysts) to $91.9bn (moomoo) to $92.07bn (Motley Fool). EPS estimates range $2.01 (TipRanks-sourced) to $2.08–$2.09. I have used $91.9bn and $2.08. Confirm against a Bloomberg or LSEG terminal pull before filing.
- Prior-year comparison. Q2 FY26 revenue $46.74bn and EPS $1.05 are consistent across sources, giving roughly 95–97% growth. One source cites $0.99 as the prior-year EPS base; unverified.
- Q1 FY27 data centre revenue. Reported as $75.246bn in most sources, but one outlet (Intellectia) states $71.6bn. I have used $75.2bn, which matches the SEC filing references. Verify.
- $105bn OpenAI lease guarantee. Sourced to a securities filing via secondary reporting (Coinotag). Separate reports (WSJ, Bloomberg, July 2026) describe discussions over a $250bn guarantee and $350bn of GPU financing, which may not have been executed. The $105bn figure needs primary 10-Q/8-K confirmation before publication.
- Supply commitments. Sources give $119bn (24/7 Wall St) and $145bn+ (Hudson Labs) for purchase commitments. I have omitted the figure rather than pick one. Worth adding once verified.
- Analyst coverage. "59 of 62 buy or higher" is from a Stocktwits piece dated 24 August. TipRanks separately shows 26 of 26 buy with a $301.82 average target. Cantor Fitzgerald has $350. Consensus target described as 40–46% above last close, depending on source and date. I have avoided quoting a specific target number.
- Share price and market cap. $210 and $5.2trn are correct as of Monday 24 August close. Both will move before publication. Year-to-date performance is inconsistent across sources: 2% (early August), 11% (Forbes, 24 August), 16% (Crypto Basic, 21 August). I have omitted YTD for that reason.
- Options implied move. 5–6% range drawn from two sources (TradingKey, ts2.tech). The $313bn market-value figure comes from ts2.tech and is a derived calculation, not an exchange figure. I have paraphrased it as "more than $300bn" in the standfirst.
- Post-earnings drop average. 2.79% one-day average across the last four reports is from Coinotag. Options AI data shows -1.77%, -5.46%, -3.15%, -0.79% for the last four, which averages 2.79%. Consistent.
- $105bn appears twice in the piece in different contexts (the Q3 guide bar is $105bn, the OpenAI guarantee is $105bn). Coincidental but potentially confusing to readers. Flagging for a possible edit.
- PCE release. Scheduled for 26 August per Finance Calendar. Confirm against the BEA calendar.
- moomoo attribution. The $105bn Q3 bar is a moomoo community analysis, not a sell-side house view. Consider replacing with a named broker if one can be sourced, as moomoo is a weaker attribution for a Substack piece.
