Meta and AMD Announce Multi-Year AI Infrastructure Agreement Up to 6GW

A Large-Scale Infrastructure Deal for the AI Era

Meta announced on February 24, 2026 that it has signed a multi-year agreement with AMD to power its AI infrastructure with up to 6GW of AMD Instinct GPUs. The company frames the partnership as a core building block for scaling compute needed for next-generation models and what it calls “personal superintelligence.”

More Than a GPU Procurement Contract

According to Meta’s announcement, the partnership is not limited to chip supply. It includes alignment across silicon, systems, and software roadmaps, aiming at tighter vertical integration across Meta’s infrastructure stack. AMD characterized the collaboration as multi-year and multi-generation, spanning Instinct GPUs, EPYC CPUs, and rack-scale AI systems.

Meta also stated that shipments supporting first GPU deployments are expected to begin in the second half of 2026. Those initial deployments are planned around the Helios rack-scale architecture, which Meta says was developed and announced with AMD at last year’s Open Compute Project Global Summit.

Portfolio Compute Strategy: Diversification + In-House Silicon

Meta places the AMD agreement inside its broader “Meta Compute” strategy. The explicit objective is to build a flexible and resilient infrastructure portfolio by combining hardware from multiple partners with its in-house MTIA (Meta Training and Inference Accelerator) silicon program. In other words, rather than relying on a single vendor path, Meta is building a diversified capacity model to match heterogeneous AI workloads across training and inference.

Why This Matters

The announcement highlights how frontier AI competition is increasingly infrastructure-bound. Model capability gains are now coupled to access to power, chips, rack architecture, networking, and software optimization at very large scale. A 6GW target signals a long-horizon deployment strategy rather than a short-term procurement move.

For Meta, the deal supports compute diversification and potentially better inference economics at scale. For AMD, it places its Instinct and EPYC platforms deeper inside one of the largest AI infrastructure expansions currently underway. Strategically, both companies are aligning roadmaps early to reduce integration friction as deployment volumes rise.

Reference: Meta Newsroom: Meta and AMD Partner for Longterm AI Infrastructure Agreement