Advanced Micro Devices (AMD) reported record Q3 2025 results with 36% revenue growth and all-time high margins. But the real story isn't in the numbers—it's what happens next.
OpenAI's "well over $100 billion" commitment to AMD Instinct GPUs represents the largest AI infrastructure bet outside NVIDIA in history. This isn't another cloud customer kicking the tires on MI300 for inference workloads. This is OpenAI—the company training GPT-5 and defining the frontier of AI—betting six gigawatts of compute on AMD for large-scale model training starting H2 2026.
Why this changes everything: The deal validates that AMD has broken NVIDIA's stranglehold on AI training. For years, the narrative was "NVIDIA for training, AMD for inference." OpenAI just obliterated that distinction.
If the world's most demanding AI workloads can run on MI450 at scale, every hyperscaler suddenly has a credible Plan B—and a 30-40% cost advantage.
The Setup is Asymmetric
Our scenario analysis shows 70% probability of +12% to +80% returns versus 30% probability of -44% to -74% losses. Expected value: +10.7%. The stock trades at 47x 2026E P/E, pricing in significant success but not full bull case. If MI450 launches on time and OpenAI's first gigawatt goes live in H2 2026, AMD could capture 20%+ of the AI GPU market by 2027—driving revenue from $25.8B (2024) to $48.5B (2027E) with EBIT margins expanding from 8% to 28%.
But Execution Risk is Real
This thesis lives or dies on three milestones: (1) MI450 launching H2 2026 without delays, (2) OpenAI's deployment starting on schedule, and (3) ROCm software reaching 80%+ CUDA parity. Miss any of these, and the bear case ($143 target, -44%) is in play. We recommend a tranched 6-10% portfolio position, building over 6-12 months as catalysts de-risk.
Download the full 20-slide analysis below for our complete financial model, scenario drivers, quarterly monitoring checklist, and position sizing framework.
