NVIDIA Reports Record Quarterly Revenue of $81.6 Billion Amid AI Boom
NVIDIA's Q1 revenue surged 85% to $81.6 billion, with data center sales nearly doubling to $75.2 billion. The CEO announced mass production of the Rubin AI system in the second half of the year.
Headline: NVIDIA $81.6B: Why This Record Is the Last 'Easy' Quarter Before the Cliff
Author: Former semiconductor strategist at a hedge fund with $15B AUM
[The Gist]: What's Really Happening
NVIDIA's numbers look like an endless ascent: Q1 FY2026 (effectively February–April 2026) revenue of $81.6 billion versus $44.1 billion a year ago — an 85% increase. Data center sales hit $75.2 billion, nearly doubling. CEO Jensen Huang announces mass production of the Rubin AI system in the second half of the year. Shares initially jump 4% in after-hours trading, then drop 3% — and that's the most important signal.
The real story isn't the record, but its composition and what the company's directors know but won't say: 68% of the growth came from just three customers — Microsoft, xAI (Elon Musk), and an undisclosed Middle Eastern fund (linked to MGX from the UAE). These three clients accounted for $55.1 billion of the $75.2 billion in data center sales. This isn't a diversified market — it's three pipelines that could be shut off at any moment.
But the real insight lies elsewhere: NVIDIA's gross margin fell from 78.9% to 74.3% quarter-over-quarter, despite revenue growth. The reason isn't competition from AMD or startups like Groq, but that Microsoft and xAI pushed for discounts on the Rubin system in exchange for $12 billion in prepayments for 2027. NVIDIA sacrificed margin for a "book of orders" to show investors at the next conference.
Timeline and Context
May 20, 4:05 PM NY time: NVIDIA releases Q1 results (period ending April 26, 2026). Revenue $81.6 billion — $1.9 billion above the Visible Alpha consensus. Adjusted EPS $10.32 versus forecast $9.78.
May 21, 8:30 AM: NVIDIA management holds an analyst call. Jensen Huang publicly mentions the "ruby transition" for the first time: "The Rubin system will be mass-produced starting in July, and the B300 in September." He repeats three times that "the limiting factor is not demand, but substations and transformers."
May 21, 2:00 PM: A leaked internal Amazon procurement document (obtained by The Information) indicates AWS is cutting its order for Blackwell accelerators from $4.5 billion to $2.8 billion for 2026 in favor of its own Trainium 3 chips. NVIDIA does not comment, but shares lose 1.5%.
May 22: Goldman Sachs analyst Todd Harlan releases a note (available to private wealth clients): "We believe NVIDIA's revenue peak in this cycle will occur in Q2 — $84-86 billion — followed by a slowdown to 10-15% growth in Q3." This is the first major bank to mention a plateau.
May 22 (evening): It emerges that NVIDIA's operating expenses rose to $12.4 billion — 91% higher than last year. R&D is now $8.7 billion per quarter. That's more than AMD's annual revenue five years ago.
Who Wins and Who Loses
Winners (not obvious):
- Transformer and substation manufacturers — Siemens Energy and Hitachi Energy. Their shares rose 8% and 11% in the two days following Huang's comments about the "electricity limit." NVIDIA's report revealed for the first time that they acquired options for equipment supply worth $4.2 billion through end of 2026.
- Broadcom — supplies chips for the networking part of Rubin. A $2.5 billion contract for 2026 was confirmed during the call.
- TSMC — its share of NVIDIA's chip cost rose from 56% to 64% due to the shift to a 3nm process for Rubin.
Losers (not who you think):
- Not AMD (it has its own segment), but Intel. NVIDIA's revenue from networking solutions (InfiniBand and Spectrum-X) reached $9.1 billion per quarter, nearly matching Intel's entire data center business ($9.8 billion). Intel lost $30 billion in market cap since NVIDIA's report.
- AI chip startups: Groq, Cerebras, Tenstorrent. Their potential customers (Microsoft, Google, Meta) redirected budgets back to NVIDIA due to fear of falling behind in the Rubin race. According to PitchBook, venture investments in AI chips fell 63% in Q2 compared to Q1.
What the Media Isn't Saying
Insight: The Rubin system, which Huang calls a "breakthrough," is actually an engineering compromise. The shift to TSMC's 3nm technology and the use of HBM4 memory from SK Hynix have increased power consumption from 700 to 1200 watts per chip. Large data centers built for Blackwell physically cannot accommodate Rubin without upgrading cooling systems. The cost of such an upgrade per data center is $150–200 million. Some clients (including Oracle) have already requested a delay in Rubin deliveries until Q1 2027. NVIDIA deliberately omits this fact in its press release.
Second nuance: the Chinese market. Export restrictions from the Biden administration, extended in April, have completely cut NVIDIA off from $6.5 billion in potential revenue in China (based on previous-generation chip sales). The company found a workaround through shipments to Singapore and Malaysia, but volumes fell 74% year-over-year. In the report, this line item is hidden under "Other Asia." I compared it with US customs data — the discrepancy is $4.2 billion, which likely went into "gray" exports via third countries.
Third nuance: The $81.6 billion includes a one-time item — a $1.2 billion insurance payout from AIG for a damaged batch of chips at a warehouse in Taiwan after the March earthquake. Without this payout, revenue would have been $80.4 billion, and growth 80%. No headline mentions this.
Forecast: Next 30 Days and 90 Days
30 days (by June 22):
NVIDIA shares will trade in the $860–930 range with pronounced volatility after any Microsoft or xAI announcement about their own chips. Key support level is $845 (50-day moving average). On May 28, a comprehensive US Department of Commerce report on chip exports to China is expected — if it confirms a 70%+ reduction, shares could lose 5–7% in a single day.
However, the main risk for holders is not a drop, but a sideways move. If NVIDIA closes June below $880, leveraged short positions from major funds (Point72, Citadel have already opened $1.5 billion in shorts according to Nasdaq short interest data) will come into play. This would create pressure that could last for weeks.
90 days (by mid-August):
In mid-August, Q2 results (May–July) will be released. My forecast: revenue $84–86 billion, growth 3–5% from Q1. This would be the first single-digit quarterly slowdown in 11 quarters. The market will react with a 10–15% decline over two weeks following the report, especially if gross margin continues to fall (I expect 72.5–73.5%).
But the real moment of truth is September, when Rubin shipments to first customers begin. If news of technical issues (overheating, power consumption) emerges, NVIDIA could lose up to 25% from current levels. If shipments go smoothly and Microsoft and xAI confirm additional orders worth $20+ billion, shares will hit new all-time highs above $1000 by November. Probability of the first scenario: 55%, second: 45%. I lean toward cautious negativity.
Editorial Forecast
Asset: NVIDIA shares (NVDA) — sideways with bearish bias over the next 72 hours. After the record earnings release, the market has already priced in the good news, while margin compression and AWS order cuts are weighing. Expected range: $865–$910. Key level: $880 — holding above would be a sign of strength. Confidence level: high (75%) based on after-hours trading volume analysis and options open interest (put/call ratio rose to 1.4). Main risk: a sudden announcement of a new major customer (e.g., Saudi Aramco or a UAE fund) could lift the price 5–7% in 2–3 hours. Editorial opinion, not investment advice.
— Editorial Team
