Meta and AMD Extend Partnership to Power AI Infrastructure with Up to 6GW of Instinct GPUs
Meta has unveiled a multi-year agreement with AMD to scale its AI infrastructure using as much as 6 gigawatts of AMD Instinct GPUs. The collaboration aims to support Meta’s vision of “personal superintelligence” by delivering massive, scalable compute power capable of handling the evolving demands of large-scale AI workloads.
Strengthening a Key Industry Collaboration Under the new deal, Meta and AMD will coordinate on silicon, systems, and software roadmaps to enable deeper vertical integration across the entire infrastructure stack. This aligned effort is designed to accelerate innovation by smoothing the handoffs between hardware and software, allowing the teams to move faster at scale.
Dr. Lisa Su, AMD chair and CEO, emphasized the strategic importance of the expanded partnership, highlighting the joint focus on high-performance, energy-efficient infrastructure tailored to Meta’s workloads. The initiative covers multiple generations of Instinct GPUs, EPYC CPUs, and rack-scale AI systems, with shipments slated to begin in the second half of 2026. The deployments will be built on the Helios rack-scale architecture, which was introduced in collaboration with AMD at the Open Compute Project Global Summit last year.
Mark Zuckerberg, Meta founder and CEO, framed the collaboration as a meaningful step toward diversifying Meta’s compute foundation and enabling efficient inference workloads. He noted that AMD is poised to play a significant role for many years as Meta expands its AI capabilities.
A Portfolio-Based Approach to AI Compute The agreement is part of Meta Compute, the company’s broader effort to scale infrastructure for a world of personal AI assistants. By diversifying partners and technologies, Meta aims to build a more resilient and flexible platform. The strategy combines hardware from a range of suppliers with Meta’s own MTIA (Meta Training and Inference Accelerator) silicon program, supporting both training and inference at scale.
This portfolio approach is intended to accelerate delivery of powerful, energy-efficient hardware tightly integrated with Meta’s software stack, enabling rapid expansion of AI capabilities to billions of people worldwide.
Forward-looking statements notice This post contains forward-looking statements about Meta’s business and roadmap. These statements involve risks and uncertainties and are not guarantees of future performance. For additional information, please review Meta’s latest Form 10-K filed with the Securities and Exchange Commission. Meta does not undertake an obligation to update these statements as a result of new information or future events.