Four days ago I wrote that May 20 would test both ends of the AI hardware chain. It did. NVIDIA delivered the most dominant quarter in semiconductor history. The stock fell.
The Numbers
Revenue: $81.6 billion. Consensus was $79.2B. Beat by $2.4B, or 3.1%. Year-over-year growth: 85%. Data Center alone: $75.25B, up 92% YoY — now 92% of total revenue.
EPS: $1.87 non-GAAP, versus $1.76–$1.78 consensus. Beat by 6.25%.
Gross margin: 75.0% non-GAAP. In line with the company's own 75% guide, above the 74.5% Street consensus. Up from 60.8% a year ago. Stable from Q4.
These are not normal numbers. $81.6B in quarterly revenue exceeds NVIDIA's entire fiscal year 2024 ($60.9B). One quarter now outpaces what was, 24 months ago, a record-breaking full year.
The Guide That Broke the Whisper
$91 billion. Plus or minus 2%. The Street expected $85–87B. The whisper was $87–90B. NVIDIA didn't just clear the whisper — it rendered it conservative by $1–4 billion. Q2 gross margin guided to 74.9–75.0%, essentially flat. No margin compression. No pricing pressure. Just acceleration.
Jensen Huang on the call: "The buildout of AI factories — the largest infrastructure expansion in human history — is accelerating at extraordinary speed." He said hyperscale capex is at $1 trillion this year, and NVIDIA should be growing faster than that spend. CFO Colette Kress introduced the Vera CPU as opening "a brand new $200 billion tab" — every major hyperscaler partnering to deploy it.
The Four Questions, Answered
In my preview, I laid out four questions. Here's where they landed:
My framework: below $79B disappoints, $80B+ clears, $81B+ = upside. Result: upside.
My framework: $85–87B = Street, $87–90B = whisper cleared, $90B+ = blowout. Result: blowout.
My framework: below 73% = pricing pressure. Result: met guide, above consensus. Q2 guide 74.9–75.0%. No compression.
Jensen called Vera Rubin "off to a tremendous start." A $91B guide issued the day before Samsung's 47,000-worker strike begins implies zero supply panic. SK Hynix is absorbing, or the guide has cushion, or both.
Four questions. Four clears. The capex chain stands at 9/9 — demand confirmed from cloud to silicon, with the anchor delivering beyond even the optimistic case.
And the Stock Fell
NVIDIA closed the regular session at roughly $222. After hours, following the press release and earnings call, the stock traded down to approximately $218 — a decline of about 2% on what should be, by any historical standard, a blowout quarter.
The beat-and-sell pattern I've tracked across five quarters now looks like this:
| Quarter | Rev Beat | Guide vs Street | Stock Reaction |
|---|---|---|---|
| Q1 FY2027 (TODAY) | +3.1% | +$4–6B above | ~-2% AH |
| Q4 FY2026 (Feb '26) | +3.4% | Above | -5.5% |
| Q3 FY2026 (Nov '25) | +3.9% | Above | -3.2% |
| Q2 FY2026 (Aug '25) | +4.1% | Above | -6.4% |
| Q1 FY2026 (May '25) | +5.6% | Above | +2.1% |
Five consecutive beats. Four sells. The only quarter that rallied (Q1 FY2026, May '25) had the largest beat at 5.6% — and that was a year ago, before the market learned to front-run the beat.
What's different this time: the $91B guide is the strongest forward signal NVIDIA has ever given. Revenue acceleration from $81.6B to $91B implies 11.5% sequential growth. The $80B buyback authorization and 25x dividend increase ($0.01 to $0.25/share) add capital return signals the market usually rewards. Yet after hours, the stock drifted lower.
What This Tells You
The market isn't saying NVIDIA is wrong. It's saying NVIDIA is known. A $5.5 trillion company guiding $91B in quarterly revenue is the most consensus long in the world. When 57 analysts rate it Buy and the average price target is $269, the information edge from beating estimates approaches zero. The beat is priced before it's reported.
The buried signal isn't in the quarter — it's in the Vera CPU announcement. Jensen positioning NVIDIA as "the world's leading CPU supplier" with a $200B TAM expansion is the thesis that could re-rate the stock. But that's a 2027–2028 revenue story, not a Q2 FY2027 story. The market doesn't pay today for revenue it can't model for two years.
Samsung's shadow matters too. The $91B guide issued hours before 47,000 Samsung workers walk out implies either ironclad SK Hynix supply, sufficient guide cushion to absorb disruption, or Jensen's confidence that the strike resolves. If it doesn't resolve and HBM4 yield stabilization gets disrupted, the $91B becomes the ceiling rather than the floor.
The Verdict
NVIDIA is the best-executing company in the world right now. $81.6B revenue, $91B guide, 75% gross margins, $80B buyback. The capex chain thesis is confirmed: 9/9, demand anchor validated with room to spare. The business is flawless.
The stock is a different question. NVDA trades at roughly 30x next-twelve-months earnings on numbers that assume the $91B guide is a floor, not a ceiling. At this valuation, NVIDIA doesn't need to be great — it needs to be greater than great, faster than expected. Today it was exactly great enough, and the stock fell.
My position: NVDA sells off 3–6% by Friday as the beat-and-sell pattern completes for the fifth time in six quarters. The $91B guide prevents a deeper rout, but doesn't break the pattern. The trade is: buy the post-earnings dip below $210 for the Q2 ramp. The next catalyst that actually moves the multiple is the first $95B+ guide — probably Q2 results in August. Until then, the bar has simply moved again.