OpenAI and AMD have cut chip design timelines from a year to under five months by using AI models to rewrite SystemVerilog RTL code for AMD's flagship GPU architecture. Verified independently by AMD, the AI-driven modifications achieved up to 14% improvements in power, performance, and area. AMD has already incorporated model-generated changes into a recently taped-out design sent to manufacturing as part of its multi-year partnership with OpenAI.
OpenAI optimization of AMD GPU design
- ▪OpenAI's optimization of AMD's GPU architecture achieved up to 14% improvements in power, performance, and area (PPA) over baseline designs
- ▪OpenAI used its AI models to rewrite SystemVerilog Register Transfer Level (RTL) code to explore alternative circuit implementations and pipeline arrangements for AMD's GPU architecture
OpenAI and AMD hardware deployment agreement
- ▪OpenAI plans to begin bringing AMD Helios rack-scale systems online in the fourth quarter of 2026, with hardware deployments accelerating through 2027
- ▪AMD issued OpenAI a warrant for up to 160 million shares, with vesting tied to deployment, technical, and commercial milestones under their multi-year partnership
- ▪Under a multi-year partnership, OpenAI committed to deploying up to 6 gigawatts of AMD Instinct GPUs, starting with an initial 1 GW deployment of MI450 GPUs in the second half of 2026
Story comments
Loading comments…