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.