AMD reported Q1 2026 after the close tonight. Every number that analysts cared about came in above expectations. Then the company guided Q2 to $11.2 billion — seven hundred million dollars above what the Street was modeling.
The stock fell 5%.
This is the fifth consecutive earnings report where AMD has beaten and the stock has sold off. The average post-earnings move across those five prints is −5.15%. Tonight’s −5.27% is statistically identical to the pattern. The market isn’t reacting to these results. It’s reacting to the last sixty-six percent.
The Scorecard
| Metric | Estimate | Actual | Beat |
|---|---|---|---|
| Revenue | $9.85B | $10.25B | +$400M / +4.1% |
| Non-GAAP EPS | $1.27 | $1.37 | +$0.10 / +7.9% |
| Data Center revenue | $5.56B | $5.78B | +$220M / +3.9% |
| Non-GAAP gross margin | 55% | 55% | In line |
| Q2 revenue guide | $10.50B | $11.2B | +$700M / +6.7% |
| Q2 gross margin guide | — | ~56% | Improving |
Five green, one amber. Revenue beat. Earnings beat. Data center beat. Guidance beat — by a wider margin than the quarter itself. The only metric that didn’t exceed expectations was gross margin, which landed exactly where management said it would.
Look at that last row. Q2 non-GAAP gross margin guided to 56%, up from 55% in Q1. Revenue accelerating and margins expanding. This is the opposite of what the bear case predicted.
The China Subtraction
The number buried inside the data center line is the one worth studying.
In Q4 2025, AMD sold approximately $390 million of MI308 GPUs into China. In Q1 2026, that number collapsed to approximately $100 million. Export controls under Section 232 effectively removed $290 million of quarterly revenue — an annualized hit of $1.16 billion.
Data center grew $2.1 billion year-over-year while losing $290 million of sequential China revenue. The non-China data center business had to grow by approximately $2.4 billion — roughly 73% — just to net out to 57% total growth. Export controls were supposed to be AMD’s structural vulnerability. Instead, hyperscaler demand in the US and allied markets absorbed the entire loss and then some.
Lisa Su declined to forecast any additional China GPU revenue beyond the ~$100 million run rate. License applications for MI325 are pending. If one gets approved in H2, it’s pure upside against a base that already prices in near-zero China.
The Rest of the Business
| Segment | Revenue | YoY Growth |
|---|---|---|
| Data Center | $5.78B | +57% |
| Client (Ryzen) | $2.89B | +26% |
| Gaming | $720M | +11% |
| Embedded | $873M | +6% |
Every segment grew. Client is benefiting from AI PC demand and Ryzen market share gains against Intel. Gaming recovered modestly. Embedded — AMD’s sleepiest segment — posted its first meaningful growth in several quarters. But these are supporting actors. Data center is 56% of total revenue now and growing three times faster than the next-largest segment.
The Cash Machine
A number that nobody will headline tonight: free cash flow was $2.57 billion.
In Q1 2025, it was $727 million. That’s a 253% increase — 3.5 times the prior year’s level. Annualized, AMD is now generating over $10 billion in free cash flow. The company holds $12.35 billion in cash and short-term investments against $3.22 billion in debt. Net cash position: $9.1 billion.
AMD was historically a revenue growth story with questionable cash conversion. That narrative is dead. At $10 billion+ annualized FCF and a $575 billion market cap, AMD trades at roughly 57 times free cash flow. Expensive — but the denominator is growing at 253% and the company now has more cash than it knows what to do with. Share buybacks were only $221 million this quarter. That number will go up.
Link Seven
I have tracked the AI capex supercycle through its physical supply chain this earnings season. AMD is the seventh link:
From lithography to fabrication to foundry proof to steel to cloud infrastructure to AI silicon — seven consecutive links have beaten and/or raised this quarter. ARM reports Wednesday. If it confirms, the entire physical supply chain of the AI capex supercycle will have cleared Q1 without a single miss.
Meanwhile, PLTR reported yesterday: US commercial revenue +133%, adjusted operating margin 60%. That’s the software layer confirming what the hardware layer just told us. Enterprise AI spend is accelerating, not plateauing.
The MI450 Signal
Lisa Su’s most important sentence tonight was about the future, not the quarter: “Customer engagement around MI450 Series and Helios is strengthening, with leading customer forecasts exceeding our initial expectations and a growing pipeline of large-scale deployments.”
The MI450 is AMD’s next-generation AI GPU, with revenue starting in Q3 and ramping to significant volume in Q4. The multi-gigawatt deals are already signed: OpenAI committed to 6GW of AMD Instinct GPUs with the first 1GW of MI450s deploying in H2 2026. Meta signed a $60 billion, five-year agreement for MI450 GPUs. Eight of the top ten AI companies now run Instinct accelerators in production.
These are not speculative letters of intent. These are infrastructure contracts for power and space that take eighteen months to build out. When Su says “forecasts exceeding initial expectations,” she is saying the customers are pulling MI450 volume forward.
Verdict
The −5% is a reflex, not an analysis.
AMD has now demonstrated that AI demand is strong enough to overwhelm three simultaneous headwinds: a $1.16 billion annualized China export control hit, memory cost inflation across the HBM supply chain, and the early-ramp margin pressure of a new product cycle (MI350). Revenue beat. Earnings beat. Guidance beat by a wider margin than the quarter. Margins held exactly at guide and are set to improve. Free cash flow tripled. And the stock fell the same 5% it falls every quarter regardless of results.
At 65 times forward non-GAAP earnings, AMD needs to grow into its valuation. The Q2 guide of $11.2 billion — implying 46% year-over-year growth and accelerating from Q1’s 38% — says it will. If the MI450 ramp in H2 delivers what the customer forecasts suggest, AMD’s exit rate for 2026 could approach $50 billion annualized. That would put forward earnings closer to $7+, and the current $360 price at ~51 times. Expensive, but less expensive than it looks tonight.
The question entering this print was whether AMD could grow through China, through cost inflation, and through the expectations embedded in a +66% YTD stock. The answer was yes, yes, and yes. The market answered with the same sell program it runs every quarter. Eventually, the forward numbers will be too large to fade.
Eleven point two billion is the number the Street wasn’t modeling. That’s where this goes next.
AMD (AMD) Q1 2026. Non-GAAP EPS $1.37 vs $1.27 est (+7.9%). Revenue $10.25B vs $9.85B est (+4.1%). Data Center $5.78B (+57% YoY). Q2 guide $11.2B vs $10.5B est (+6.7%). GAAP gross margin 53%, non-GAAP 55%. FCF $2.57B (+253% YoY). Stock −5.3% AH. Cathie Wood sold $79.9M pre-earnings. Brent $110. S&P 500 7,230.