Meta inks multi-year AMD deal for over 6GW of AI GPU capacity
Meta has inked a multi-year deal with AMD to deploy over 6 gigawatts of AI GPU capacity across its data centers. The agreement represents AMD’s largest hyperscale cloud partnership for its CDNA3 architecture and bolsters its data center revenue outlook through 2028.
1. Meta-AMD Supply Agreement
Meta has entered a multi-year contract with AMD to deliver over 6 gigawatts of AI GPU capacity to its data centers, marking AMD’s largest hyperscale cloud partnership to date. Hardware deliveries will commence in the second half of 2026 and scale to full capacity by 2028.
2. CDNA3 Architecture Deployment
The GPUs leverage AMD’s third-generation CDNA architecture, designed for large-scale training and inference workloads. Each accelerator draws up to 600 watts of power, offering improved performance-per-watt and enabling Meta to run more complex AI models on-premises.
3. Revenue and Competitive Implications
The agreement is set to drive double-digit annual growth in AMD’s data center segment revenue through 2028, potentially adding over $2 billion in high-margin sales. It also challenges Nvidia’s dominance in hyperscale AI compute by expanding AMD’s footprint among top cloud providers.