AMD and Meta Establish Historic 6 Gigawatt AI Infrastructure Agreement
AMD and Meta established a historic agreement which enables them to develop AI infrastructure with a capacity of 6 gigawatts. The agreement establishes a massive 6 gigawatt deployment of AMD Instinct GPUs and high performance CPUs which will serve as the foundation for the companies expanded strategic relationship.
The partnership develops hardware which Meta uses to operate its AI workloads. The deployment consists of these main components
- Custom Instinct MI450 GPUs These GPUs rated for gigawatt level AI efficiency use the MI450 design as their foundation. The first shipments of the product will start delivering in the second half of 2026.
- AMD Helios Rack Scale Architecture The Open Compute Project serves as the platform through which AMD and Meta jointly developed Helios which delivers scalable infrastructure to support AI services across the world.
- 6th Gen AMD EPYC CPUs The processors with codenames of Venice and Verano will operate the AI stack through their orchestration and general computing functions while delivering power efficient performance.
- ROCm Software The system will depend on the open source ROCm software stack to provide smooth GPU and CPU connection capabilities.
The agreement establishes a performance based incentive system which will link the strategic goals of both companies. AMD has issued Meta warrants to purchase up to 160 million shares of AMD common stock. The company will grant warrants which depend on achievement of particular milestones
- The first portion vests upon the initial 1 gigawatt shipment of Instinct GPUs.
- The next portions will become available to Meta after it achieves the 6 gigawatt purchasing goal.
The parties will determine the process of vesting and exercise based on AMD reaching stock price milestones and Meta delivering on its commercial and technical objectives.
The agreement establishes AMD EPYC processors as a critical component of Meta's international data center operations which go beyond GPU technology in their support. Meta operates already with hundreds of thousands of EPYC CPUs and earlier MI300 and MI350 GPU products. Through its position as lead customer for 6th Gen Venice architecture Meta will gain access to new computing resources which will help it maintain its position as energy efficient AI inference leader in the field.
The two companies see this agreement as essential for developing personal superintelligence which will strengthen human connections through AI interactions. The partnership will drive long term revenue growth for AMD which will establish the company as a leading supplier for worldwide AI infrastructure development projects. Meta requires specific silicon which has capacity for millions of active users to support AI operations across its platforms of Instagram, WhatsApp, and Messenger.
