Meta is aggressively diversifying its AI hardware supply chain, striking a monumental deal with AMD worth up to $100 billion. This multiyear agreement is a cornerstone of Meta’s strategy to build out the infrastructure needed for “personal superintelligence”—AI systems designed to understand and assist users on a deeply individual level.
The Logistics of the Deal
The partnership involves Meta purchasing a massive volume of AMD’s MI540 series GPUs and its latest generation of CPUs. This hardware influx is expected to drive approximately six gigawatts of data center power demand.
To solidify the partnership, AMD has issued Meta performance-based warrants for up to 160 million shares of common stock—roughly 10% of the company—at a nominal price of $0.01 per share. These awards are tied to specific milestones, including a requirement for AMD’s stock price to reach $600 for the final tranche to vest. For context, AMD shares recently closed near $196.
Shifting Away from Nvidia Dominance
While Nvidia remains the dominant force in the AI sector, Meta’s move signals a broader industry trend toward diversification. By integrating AMD’s hardware, Meta reduces its reliance on a single vendor and leverages the growing efficiency of CPUs for AI inference and agentic scaling.
AMD CEO Lisa Su noted that the CPU market is seeing explosive growth as infrastructure deployments scale. This deal mirrors a similar equity-for-hardware agreement struck between OpenAI and AMD.
Infrastructure and the Road Ahead
Meta’s hardware spree is part of a staggering $600 billion commitment to U.S. data centers and AI infrastructure over several years. For 2026 alone, the company projects $135 billion in capital expenditures. Key projects include a $10 billion gas-powered data center campus in Indiana.
Although Meta continues to work with Nvidia and develop its own in-house silicon, the AMD deal provides the immediate, high-scale compute necessary to fuel its long-term vision for agentic AI.
