Advanced Micro Devices (AMD) has agreed a multi year partnership with Meta to build next generation artificial intelligence infrastructure, under which AMD will deploy up to six gigawatts (GW) of its AMD Instinct graphics processing units (GPUs) to support Meta's expanding AI operations. The companies said the deal strengthens their strategic partnership and aligns roadmaps across silicon, systems and software as they aim to deliver platforms tailored to Meta's large scale workloads.

The first phase will use a custom AMD Instinct GPU based on the MI450 architecture and shipments for the initial one gigawatt deployment are expected to begin in the second half of 2026. Systems will also include sixth generation AMD EPYC central processing units (CPUs) codenamed Venice and will run on AMD's ROCm software platform to manage and optimise large scale model training and inference workloads.

The infrastructure will be built on the AMD Helios rack scale architecture that was jointly developed with Meta through the Open Compute Project to support scalable AI infrastructure. Meta will be a lead customer for AMD's upcoming sixth generation EPYC processors, including Venice and Verano, which are designed to deliver strong performance while improving cost and energy efficiency as the company diversifies its compute base.

As part of the commercial arrangement AMD has issued Meta a performance based warrant for up to 160 million shares of AMD common stock that will vest in stages as shipment milestones are met and purchases scale from the first one gigawatt deployment to six gigawatts, with vesting also linked to stock price targets and technical and commercial milestones. AMD executives said the partnership is expected to drive significant multi year revenue growth and strengthen long term financial performance given Meta's prior deployments of EPYC CPUs and large volumes of Instinct MI300 and MI350 series GPUs across global data centres.