Nvidia Before August 26: One Report to Settle the AI Debate
By Ieva Bružaitė | Founder, Orion Capital | August 2026
Executive Summary
Markets have spent the entire summer arguing about artificial intelligence — through a June selloff, a July bear market in semiconductors, a violent earnings-week reversal and a Palantir quarter that grew 93%. Every one of those episodes was a proxy battle. The actual verdict arrives on Wednesday, August 26, after the close, when Nvidia reports the quarter ended in late July.
The strange part is the setup. The company that created the AI trade has become its laggard: the shares sit around $193, roughly 18% below May’s all-time high, and have returned only about 20% over twelve months — trailing the very technology sector their products power. Apple, not Nvidia, reclaimed the title of the world’s most valuable company in late July. Meanwhile the company’s own guidance calls for roughly $91 billion in quarterly revenue, a 95% increase, and forty-three of forty-seven analysts rate the stock a strong buy. A market has rarely been this bullish on paper and this nervous in price. August 26 is where the two positions reconcile.
The Setup: The Leader That Stopped Leading
Nvidia’s underperformance this year is not a mystery; it is a list. A China export overhang that refuses to resolve. Execution risk in the transition from the Blackwell architecture to next-generation Rubin. A $5 trillion valuation at which merely meeting expectations has historically been punished. And, since late July, a bond market whose 30-year yield — the highest since 2007 — quietly taxes every long-duration asset.
Yet through the same twelve months, the business itself never blinked: revenue grew 85% in the May quarter, the data centre segment grew 92% to $75.2 billion, and gross margin held near 75%. The gap between a flat stock and an accelerating income statement is the entire investment question. Either the market knows something about the future the numbers don’t show — or the numbers are about to force the market to admit it was early to worry.
What Is Actually Expected
| What | Number | Why It Matters |
| Company revenue guidance (Q2 FY27) | $91B ± 2% (~+95% YoY) | The bar Nvidia set for itself; consensus sits at ~$91.8–91.9B |
| Consensus EPS | ~$2.07 | Roughly a doubling from a year earlier |
| Prior quarter (May) | $81.6B, +85% YoY | Data centre $75.2B, +92%; gross margin ~75% |
| October-quarter guide (the real event) | Street models ~$104B | Bulls (BofA) argue for $107–108B; the guide, not the print, moves the stock |
| Blackwell FY27 revenue expectation | ~$137B (from ~$86B) | The ramp that must not stumble |
| Rubin | First revenue expected next quarter | Consensus ~$38B for FY27; every frontier lab reportedly adopting at launch |
| Stock context | ~$193, ~18% below May high | Up only ~20% in 12 months — trailing the tech sector it created |
| Analyst positioning | 43 of 47 rate strong buy; avg target ~$304 | Near-unanimity is a risk in itself |
Note the structure of the table: only two rows are about the quarter being reported. The rest are about what comes next. That is the correct weighting. After nine consecutive beats, nobody seriously doubts the $91 billion. The stock will trade on the October-quarter guide — Wall Street models roughly $104 billion; Bank of America argues management will guide $107–108 billion and open a multi-quarter upgrade cycle — and on two words: Blackwell and Rubin.
The Three Questions That Decide the Evening
First: can Blackwell ship at the pace the numbers assume? Consensus expects Blackwell revenue to rise from roughly $86 billion last fiscal year to about $137 billion this year. That is not a demand question — it is a manufacturing one, running through TSMC’s packaging lines and, critically, through High Bandwidth Memory supply.
Second: is Rubin on schedule? The next-generation platform is expected to begin contributing revenue next quarter, with consensus already pencilling in roughly $38 billion for the fiscal year. Chief executive Jensen Huang has said every frontier model company plans to adopt Rubin systems at launch — a claim that was not true of Blackwell’s debut. Any confirmed timeline, sample-shipment detail or early order commentary will be read as the report’s real headline.
Third: what does the guide say about demand durability? The October-quarter number is the market’s referendum on whether hyperscaler capital expenditure — now guided at $720–745 billion for 2026 across the four largest buyers — translates into Nvidia’s order book at the pace bulls assume, or whether custom silicon and digestion phases begin to bite.
What the Supply Chain Already Testified
Here is the advantage of covering the ecosystem rather than the ticker: most of Nvidia’s report has already been previewed, under oath, by its suppliers and customers. SK Hynix — supplier of the majority of Nvidia’s High Bandwidth Memory — reported the strongest quarter in its history in late July, confirmed that HBM4 mass shipments began in Q2 and ramp through the second half, and disclosed multi-year contracts with roughly ten customers, with capacity effectively sold out. Samsung’s report hinted at prolonged chip shortages. Korean semiconductor exports rose 179% in July. And the hyperscalers did not merely express enthusiasm — they raised written capital-expenditure guidance within days of each other.
Supply sold out, memory ramping, customers raising budgets in writing: it is difficult to assemble that evidence into a story where Nvidia’s demand quietly collapsed in the July quarter. The honest uncertainty is not whether the quarter was strong. It is whether “strong” clears a bar set by near-unanimous bullishness.
The Bear Case, Stated Fairly
- Expectations are the adversary. When 43 of 47 analysts say strong buy, meeting estimates is functionally missing them. SK Hynix just demonstrated the mechanism: record results, sold anyway, because the ruler moved faster than the business.
- The guide could disappoint by being merely good. A $104 billion October guide that matches consensus may read as deceleration against whispers of $107 billion-plus.
- China remains an unresolved discount. Export policy can subtract billions from addressable revenue with a single headline, and management can neither control nor confidently forecast it.
- Financing questions are getting louder. Nvidia’s newly reported plan to mobilise up to $500 billion of Wall Street financing for AI infrastructure invites the vendor-financing critique — that the ecosystem increasingly funds its own demand. The concern is early, but it is no longer fringe.
- Rates. A 5%-plus long bond compresses every distant cash flow, and the market has spent August re-learning that lesson weekly.
Orion Capital Perspective
Readers of this site will recognise the framework: since late July we have argued that the AI market stopped paying for promises and started paying for receipts. Microsoft and Amazon produced receipts and were rewarded; Meta produced spending and was not; Palantir produced a 93% quarter and settled the software half of the argument. August 26 is the hardware half — the final exam.
Our reading of the evidence is that the demand side of the exam is largely answered before the bell: the supply chain’s numbers, contracts and export data all point the same direction. What Nvidia’s evening will genuinely decide is narrower and, for long-term investors, more useful — the slope. Whether the AI buildout compounds from here at the pace of the bulls’ $107 billion or the base case’s $104 billion matters enormously to a stock priced at $5 trillion, and very little to the question of whether the buildout is real. Investors should therefore prepare for both outcomes of the print while refusing to confuse them: a stock that falls on a good quarter is an expectations event; a data centre number that stops growing would be a thesis event. Only the second kind should change a long-term position.
There are, as always, two clocks. The reactive one will ring loudly at 23:20 Vilnius time on August 26. The anticipatory one has been ticking steadily all summer — in Icheon’s sold-out fabs, in hyperscaler budget lines, in ten-customer memory contracts. We continue to set our watch by the second.
What We Will Watch on the Call
- The October-quarter revenue guide against the ~$104 billion consensus — the single number that moves the stock
- Any Rubin timeline specifics: sample shipments, launch-quarter revenue, early adopter commitments
- Data centre growth rate versus last quarter’s 92% — deceleration slope, not absolute level
- Gross margin trajectory through the Blackwell-to-Rubin transition
- China commentary and whether any resolution is embedded in guidance
- Any detail on the $500 billion infrastructure financing initiative and its accounting treatment
Conclusion
Every proxy battle of this summer — the selloff, the reversal, the memory war, the receipts debate — has been an argument about a report that had not yet been filed. On August 26 it will be. Our advice is the same as it has been all season: decide in advance which numbers would actually change your thesis, write them down, and let the evening’s volatility argue with someone else. The AI trade has stopped being one trade. But it still has one referee.
Disclaimer: This publication is independent research and reflects the views of the author. It is provided for informational purposes only and does not constitute investment advice or a recommendation to buy or sell any security. Investors should conduct their own analysis before making decisions.
Sources & References
- Nvidia Investors Relations – Financials Results
- Nvidia – Quarterly Results/Earnings & Guidance
- U.S SEC – NVIDIA filings
- U.S Department of Commerce/Bis – Semiconductor Export Controls
About the Author
Ieva Bružaitė is the Founder of Orion Capital, an independent market researcher covering artificial intelligence, semiconductors, technology companies and long-term investing.