Advanced Micro Devices (AMD) trades near $421 with a 52-week range of $107.67 to $469.22. The market cap is roughly $685 billion.
The article title flags a +114% gain. The verified 12-month total return is actually about +267% as of mid-May 2026, so AMD has more than doubled that headline figure.
This piece walks through what actually drove the move. Data center scale, the MI400 roadmap, and three hyperscaler deals that lock in revenue visibility into 2027.
Q1 2026 Data Center Segment: Record $5.8B Up 57% YoY
AMD reported Q1 2026 total revenue of $10.3 billion. That was up 38% year over year and ahead of consensus.
Data Center segment revenue hit a record $5.8 billion, up 57% year over year. That single segment is now more than half of total company revenue.
1. EPYC server CPUs kept compounding share
EPYC continued to take share from Intel in cloud and enterprise. Hyperscaler refresh cycles are tilting toward AMD as performance per watt favors the latest Turin generation.
Server CPU revenue is now a structural growth engine, not a cyclical print. AMD guided to roughly $120 billion in cumulative server CPU revenue through 2030.
2. Instinct GPUs scaled into AI inference
Instinct MI325X and MI300X shipments drove most of the Data Center upside in the quarter. AI workloads were heavily weighted to inference rather than training.
Non-GAAP gross margin was 55%, up one point year over year. Q2 2026 revenue guidance is $11.2 billion, plus or minus $300 million.
Why AMD Is Catching Up in AI Training and Inference
For most of 2024 and 2025, the AMD bear case was simple. CUDA had a software moat, and Instinct GPUs only mattered for inference at the margin.
That has changed in two ways. ROCm 7 closed most of the framework gap for PyTorch workloads. Memory bandwidth per dollar now favors AMD on several inference shapes.
The training story is also turning. Meta and OpenAI have both publicly committed to large training clusters on AMD silicon. That is a credibility shift more than a margin shift.
Trade AMD and other AI chip leaders on Gotrade with fractional shares from $1, 24/5 US market access, and no monthly fees.
MI400 Roadmap and Customer Wins With Microsoft, Meta, OpenAI
AMD unveiled the full MI400 series at Advancing AI 2025 and again at CES 2026. The flagship is the MI455X with 40 petaflops of FP4 compute.
Each MI455X carries 432 GB of HBM4 memory and 19.6 TB/s of bandwidth. The Helios rack-scale platform delivers up to 3 AI exaflops per rack and ships in Q3 2026.
1. The OpenAI 6 gigawatt commitment
OpenAI signed a multi-year capacity commitment for 6 gigawatts of AMD GPUs. The first 1 gigawatt deployment uses the MI450 GPU and lands in the second half of 2026.
The deal also includes a warrant for OpenAI to acquire up to 160 million AMD shares if both sides deliver. That structure aligns incentives without front-loading dilution.
2. The Meta 6 gigawatt deal worth about $60B
Meta committed to up to 6 gigawatts of custom AMD Instinct MI450 deployments. The estimated multi-year contract value is around $60 billion.
The warrant in this case could hand Meta roughly 10% of AMD's equity if delivery targets are met. Microsoft remains a Q1 commentary customer and a confirmed buyer of MI300X and MI325X capacity.
Valuation: Forward P/E vs Nvidia and Broadcom
AMD trades at roughly 58 times forward earnings as of mid-May 2026. That is a premium to the semiconductor industry median of about 37.
The comparison set tells a different story. Nvidia (NVDA) trades around 25 times forward earnings, and Broadcom (AVGO) sits near 38 times.
So AMD is the most expensive of the three on forward earnings. The bull defense is the growth rate: roughly 34% forward revenue growth versus a sector median near 10%.
The forward PEG ratio sits near 1.26 on non-GAAP earnings. That is below the sector median of 1.45, which is why the premium can be defended.
Position Sizing AMD Alongside NVDA in an AI Basket
The cleanest way to own AMD is not instead of Nvidia. It is alongside Nvidia in a basket.
Nvidia still owns the training moat and the highest margin profile in the space. AMD is the second-source insurance policy that every hyperscaler now actively wants to develop.
For wider AI semiconductor exposure, TSMC (TSM) remains the foundry that builds both. Our guide to semiconductor stocks beyond NVDA walks through the same AMD, AVGO, and TSM trio in depth.
A reasonable sizing template is roughly 50% NVDA, 25% AMD, and 25% split between AVGO and TSM. That captures the leader and the challenger without doubling exposure.
Conclusion
AMD has stopped being the cheap second pick on AI compute. It is now a $685 billion company with $5.8 billion in quarterly data center revenue and two hyperscaler power commitments that stretch into 2027.
The premium valuation is doing real work. If AMD compounds Data Center at 50%+ for another year and MI400 ships on schedule, the multiple is defensible.
If MI450 deployments slip or hyperscaler capex flattens, the 58 forward P/E compresses fast. Size it accordingly. You can trade AMD and the rest of the AI basket on Gotrade in fractional shares from $1.
FAQ
How much has AMD stock gained in the last 12 months?
AMD is up about 267% over the trailing 12 months as of mid-May 2026, well above the +114% figure in some headlines.
What was AMD's Data Center revenue in Q1 2026?
Data Center revenue was a record $5.8 billion, up 57% year over year and more than half of total company revenue.
When does the AMD MI400 series ship?
The MI455X and Helios rack-scale platform ship in Q3 2026, with MI450 hyperscaler deployments starting in the second half of 2026.
How does AMD's forward P/E compare to Nvidia?
AMD trades near 58 times forward earnings, Nvidia near 25 times, and Broadcom near 38 times in mid-May 2026.
Should I buy AMD instead of Nvidia?
Most balanced AI baskets own both. Nvidia anchors training share, while AMD adds second-source upside on inference and hyperscaler diversification.
