Meta plans to purchase potentially up to $100 billion worth of AMD chips under a new multiyear agreement announced Tuesday, a commitment large enough to drive roughly six gigawatts of data center power demand. As part of the deal, AMD issued Meta a performance-based warrant for up to 160 million shares of AMD common stock — about 10% of the company — priced at $0.01 per share, with vesting tied to specific milestones. The final tranche would require AMD’s stock to reach $600, according to The Wall Street Journal. AMD shares closed Monday at $196.60.
Under the agreement, Meta will buy AMD’s MI540 series GPUs along with its latest generation of CPUs, reflecting a broader industry shift in how AI workloads are built and scaled. While GPUs remain central to training advanced AI models, CPUs are playing an increasingly important role in inference — the process of running AI systems in production — because they’re efficient, easier to scale, and reduce reliance on Nvidia’s tightly integrated ecosystem.
“The CPU market is absolutely on fire,” AMD CEO Lisa Su said during an investor briefing Tuesday morning, pointing to surging demand driven by AI infrastructure buildouts and the growth of inferencing and agentic AI systems. She said AMD’s portfolio is well positioned as those deployments expand.
For Meta, the scale of the purchase underscores how aggressively it is investing in AI infrastructure. The company has pledged to invest at least $600 billion in U.S. data centers and AI infrastructure over the coming years, including a projected $135 billion in capital expenditures in 2026 alone. It recently unveiled plans for a $10 billion gas-powered data center campus in Indiana designed to support 1 gigawatt of compute capacity.
CEO Mark Zuckerberg described the AMD partnership as “an important step” in diversifying Meta’s compute stack as the company works toward what he calls “personal superintelligence” — AI systems designed to deeply understand and empower individuals in everyday life.
The agreement also highlights Meta’s effort to reduce its dependence on Nvidia, the dominant supplier of AI chips, which has commanded premium pricing amid overwhelming demand. Just weeks earlier, Meta struck a separate multiyear deal to expand its data centers with millions of Nvidia’s latest CPUs and GPUs. At the same time, Meta continues to develop its own in-house silicon, though those efforts have reportedly faced delays.
AMD, for its part, has been steadily gaining traction among AI developers looking for alternatives to Nvidia’s hardware. In October, AMD struck a similar equity-linked chip agreement with OpenAI, signaling that major AI firms are increasingly willing to structure long-term hardware partnerships to secure supply.
The Meta deal stands out for both its size and its structure. By tying a sizable equity award to performance milestones — and ultimately to a $600 share price target — AMD is aligning its upside with Meta’s long-term AI ambitions. If fully vested, the warrant would represent a significant ownership stake in the chipmaker.
At a moment when AI infrastructure spending is accelerating across the industry, Meta’s $100 billion commitment signals that compute capacity — not just model breakthroughs — is becoming a central competitive lever. The company isn’t abandoning Nvidia, but it is clearly broadening its supplier base as it builds out the physical foundation for its next phase of AI development.
This analysis is based on reporting from TechCrunch.
Image courtesy of AMD.
This article was generated with AI assistance and reviewed for accuracy and quality.
