Meta committed to multi-year deal purchasing AMD Instinct GPUs totaling 6GW infrastructure capacity.

Meta has announced a multi-year agreement with AMD to power its artificial intelligence infrastructure with up to 6GW of AMD Instinct GPUs. The collaboration marks a major step in Meta's long-term AI roadmap as the company accelerates efforts to build scalable infrastructure capable of delivering personal superintelligence to billions of users worldwide.

AMD will supply its Instinct GPU platforms, along with EPYC CPUs and rack-scale AI systems, to support Meta's growing AI training and inference requirements. Shipments for the first GPU deployments will begin in the second half of 2026, with the infrastructure built on the Helios rack-scale architecture, a system developed and announced at the Open Compute Project Global Summit in collaboration between Meta and AMD. Dr. Lisa Su, Chair and CEO of AMD, stated that the expanded partnership aligns both companies' roadmaps across silicon, systems and software, enabling high-performance and energy-efficient infrastructure optimized for Meta's AI workloads. Mark Zuckerberg, Founder and CEO of Meta, emphasized that the partnership represents a critical step in diversifying Meta's compute ecosystem, adding that AMD will remain an important long-term partner.

The new agreement deepens collaboration between Meta and AMD through vertical integration across silicon, systems and software layers. By aligning silicon development, rack-scale systems, and software stacks, the companies aim to enable faster innovation cycles and improved infrastructure efficiency. This integrated approach supports Meta's broader strategy to vertically optimize its AI infrastructure stack for both performance and energy efficiency at scale. The AMD partnership forms part of Meta's broader Meta Compute initiative, focused on scaling infrastructure for the era of personal superintelligence, adopting a portfolio-based approach that combines external partnerships with its in-house Meta Training and Inference Accelerator silicon program.

Alongside the AMD partnership, Meta has also expanded its multi-year strategic partnership with NVIDIA to advance its long-term AI infrastructure roadmap. According to NVIDIA Founder and CEO Jensen Huang, the collaboration includes deep co-design across CPUs, GPUs, networking and software to power one of the world's largest AI deployments. As part of the collaboration, Meta has adopted NVIDIA Confidential Computing to enhance WhatsApp private messaging security while enabling AI-powered features, NVIDIA Spectrum-X Ethernet networking platform to deliver low-latency, AI-scale networking performance, and joint optimization of AI models across Meta's core workloads to improve performance per watt and operational efficiency. With multi-generational partnerships across AMD and NVIDIA, Meta is significantly scaling its AI compute capacity, positioning the company at the forefront of large-scale AI buildouts as demand for AI-powered services continues to surge globally.