AMD Advanced Micro Devices
A real AI challenger priced near fair value; upside requires flawless MI execution at Nvidia-like margins.
The story
AMD is the credible second source in AI accelerators (MI series) and keeps gaining server CPU share with EPYC, alongside steady PC, gaming and embedded franchises. Its moat comes from chiplet design expertise, x86 duopoly economics and an improving ROCm software stack, though Nvidia's CUDA ecosystem still sets the terms. TTM revenue of $45B and an 18% operating margin show the company moving from cyclical recovery into a multi-year AI scale-up, with operating leverage still ahead.
Nothing material has changed since October: TTM revenue and margin are both moving along the path I expected, so all drivers stay the same. A 35% target margin sits between the x86 duopoly's historical economics and Nvidia's peak, and a 1.8 sales-to-capital ratio reflects a fabless model that is reinvesting heavily in R&D and acquisitions. The 12-year horizon assumes the AI-plus-datacenter runway is long but contested.
Value drivers
| Revenue growth (Y1) | 45.0% |
| Terminal growth | 4.5% |
| Forecast horizon | 12y |
| Target operating margin | 35.0% |
| Years to target margin | 6 |
| Sales-to-capital | 1.80 |
| Beta | 1.40 |
| Failure probability | 2.0% |
| Cost of capital (WACC) | 11.4% |
| Terminal WACC | 9.6% |
Valuation bridge
| PV of explicit FCFF | 391.53B |
| PV of terminal value | 751.89B |
| Equity value | 1.12T |
| ÷ shares → per share | $684.61 |
Projected free cash flow to the firm
| Yr | Revenue | Growth | Margin | NOPAT | Reinvest | FCFF | PV |
|---|---|---|---|---|---|---|---|
| 1 | 65.88B | 45.0% | 21.0% | 13.83B | 11.36B | 2.48B | 2.22B |
| 2 | 93.10B | 41.3% | 23.8% | 22.16B | 15.12B | 7.04B | 5.67B |
| 3 | 128.14B | 37.6% | 26.6% | 34.09B | 19.47B | 14.62B | 10.56B |
| 4 | 171.66B | 34.0% | 29.4% | 50.47B | 24.17B | 26.29B | 17.05B |
| 5 | 223.62B | 30.3% | 32.2% | 72.01B | 28.87B | 43.14B | 25.10B |
| 6 | 283.08B | 26.6% | 35.0% | 99.08B | 33.03B | 66.04B | 34.49B |
| 7 | 347.93B | 22.9% | 35.0% | 121.78B | 36.03B | 85.75B | 40.18B |
| 8 | 414.83B | 19.2% | 35.0% | 145.19B | 37.17B | 108.03B | 45.43B |
| 9 | 479.32B | 15.5% | 35.0% | 167.76B | 35.83B | 131.94B | 49.79B |
| 10 | 536.18B | 11.9% | 35.0% | 187.66B | 31.59B | 156.07B | 52.85B |
| 11 | 580.05B | 8.2% | 35.0% | 203.02B | 24.37B | 178.65B | 54.29B |
| 12 | 606.16B | 4.5% | 35.0% | 212.15B | 14.50B | 197.65B | 53.90B |
Key risks
- MI accelerator ramps slip or hyperscalers favor Nvidia and in-house ASICs (TPU, Trainium, MTIA)
- AI capex cycle digestion compresses datacenter GPU demand and pricing
- TSMC capacity, advanced packaging constraints, or China export controls cap shipments
Catalysts
- MI400/Helios rack-scale deployments with OpenAI and other hyperscalers converting to recognized revenue
- EPYC Venice share gains pushing server CPU share past 40%