OpenAI just locked in 6 gigawatts of AMD GPUs, a colossal bet on compute that puts another pillar under its fast-growing AI infrastructure—and signals that the race for chips now runs on multi-gigawatt commitments and creative finance.
OpenAI’s new agreement with AMD commits to a multi-year, multi-generation rollout beginning with 1 gigawatt of MI450 systems in the second half of 2026. To align incentives, AMD issued OpenAI a warrant for up to 160 million AMD shares, vesting as deployment and share-price milestones are met. For AMD, analysts frame the revenue impact in the tens of billions; for OpenAI, it’s diversified, non-NVIDIA supply at nation-scale.
Over the last 18 months, OpenAI has stitched together a web of supersized compute deals that read like an energy developer’s portfolio—only the fuel is GPUs:
- NVIDIA (letter of intent): At least 10 GW of NVIDIA systems, with NVIDIA signaling up to $100B in progressive investment as capacity is deployed; first 1 GW slated for 2H 2026 on the Vera Rubin platform. (OpenAI)
- Oracle (Stargate expansion): A landmark $300B agreement aiming to develop ~4.5 GW of additional capacity over five years, extending the Microsoft–OpenAI platform onto Oracle Cloud Infrastructure for more headroom. (OpenAI)
- Microsoft (core cloud + Stargate partner): The long-standing anchor relationship continues into the next phase, with Microsoft publicly backing the Stargate roadmap announced in 2025. (The Official Microsoft Blog)
- G42 / Stargate UAE: A 1-GW international deployment, operated with Oracle and partners, bringing nation-scale compute online in the Middle East. (OpenAI)
Why the AMD deal matters
- Supply diversification at scale. OpenAI has relied heavily on NVIDIA—now it’s adding a second top-tier supplier with multi-generation commitments, reducing single-vendor risk during a global GPU crunch.
- Financial engineering meets infrastructure. The warrant for up to 160M AMD shares ties upside to real deployments and stock targets—mirroring the performance-linked structures we’ve begun to see in other mega-deals.
- Power-first planning. Talking in gigawatts is a tell: OpenAI and its partners increasingly plan around power and site readiness first, then backfill with GPUs—a model underscored by the Oracle and NVIDIA announcements.
OpenAI’s pact with AMD doesn’t replace NVIDIA—it joins NVIDIA and Oracle (with Microsoft as the long-time cloud backbone) in a portfolio approach designed to secure >10 GW of next-gen capacity over the coming years. For the public, the translation is simple: more chips, more datacenters, and faster AI—funded by unusually large, milestone-driven deals that blur the line between tech procurement and project finance.
