Advanced Micro Devices
AMD is the clearest alternative to NVIDIA in AI infrastructure. FY2025 Q4 data center revenue hit $5.4B (+39% YoY). February 2026 AMD-Meta $60B deal signed — 6GW custom MI450 GPUs over 5 years with equity warrants. Combined with the OpenAI MI450 anchor, AMD now has two hyperscaler-committed buyers reducing CUDA lock-in exposure. Data center GPU revenue guided for 114% YoY growth to $15B in 2026.
Scenarios
Two hyperscaler anchor deals (OpenAI 1GW + Meta 6GW MI450) de-risk the 2026-2027 revenue ramp. EPYC server CPU at 25%+ market share provides a $10B+ floor even if GPU competition intensifies.
NVIDIA H200/B200 maintains 5-7x software ecosystem advantage (CUDA); AMD GPU market share is still ~15% vs NVIDIA's ~80% in AI training. If MI450 ships with performance gaps, hyperscalers can still revert to NVIDIA.
Key Factors to Watch
- ●Two hyperscaler anchors (OpenAI MI450 + Meta $60B 6GW custom MI450) validate AMD as genuine NVIDIA alternative
- ●ROCm software stack remains the key adoption barrier vs CUDA — ecosystem gap narrowing but not closed
- ●Data center GPU revenue guided for 114% YoY growth to $15B in 2026; MI400 series targets 2026-2027 deployment
Score History
| Date | Score | Direction | Note |
|---|---|---|---|
| 2026-04-10 | 7.4 | Positive | Score 7.1→7.4 (md 5→6). Meta $60B 6GW custom MI450 deal (Feb 24, 2026) + existing OpenAI anchor = two hyperscaler committed buyers reducing CUDA lock-in dependency risk. |
| 2026-03-08 | 7.1 | Positive | Score 7.5→7.1 (formula reweight: sb 0.25→0.15, are 0.20→0.25, md 0.20→0.25, dr 0.20→0.25, aam 0.15→0.10) |
| 2026-03-08 | 7.5 | Positive | Initial assessment from batch 9 blind spot review |
Software Peers
Last researched: 2026-04-25
This is research and analysis, not financial advice. Scores reflect AI impact potential, not investment recommendations.