AMD and OpenAI announce strategic partnership to deploy 6 gigawatts of AMD GPUs

OpenAI has signed a multi-year deal with AMD to deploy a staggering 6 gigawatts of Instinct GPUs, with the first gigawatt going live sometime in 2026. To put that number in perspective, a single gigawatt is roughly enough to power a mid-sized city. Six of them, dedicated purely to training and running AI models, signals just how enormous OpenAI expects its computing needs to become over the next few years.

This is as much a statement about supply chains as it is about hardware. For the last two years, Nvidia has essentially owned the AI chip market, with OpenAI, Microsoft, and nearly everyone else lining up for its GPUs amid chronic shortages. By committing to AMD at this scale, OpenAI is hedging against that dependency and giving AMD's Instinct line the kind of marquee validation it has struggled to secure. AMD has long promised its chips could rival Nvidia's on performance for large language model workloads; now it gets to prove that claim at a scale few competitors can match.

The timing also lines up with OpenAI's broader infrastructure sprint. The company has been striking data center and energy deals across the board this year, all pointing toward the same conclusion: it believes future models will need vastly more compute than what currently exists, and it wants multiple suppliers locked in before the crunch gets worse. Six gigawatts spread across a multi-year rollout gives OpenAI room to scale gradually while AMD ramps up manufacturing and delivery.

Neither company detailed the deal's dollar value, but the scale alone suggests this ranks among the largest hardware commitments in AI history. If it delivers as planned, it could reshape how the industry views AMD, not as an alternative to Nvidia, but as a legitimate second pillar of AI infrastructure.