AMD Strikes $60 Billion Multi-Year AI Chip Deal With Meta, Grants Option for Up to 10% Stake to Challenge Nvidia's Footprint

Advanced Micro Devices has secured a sweeping multi-year agreement worth up to $60 billion to supply artificial intelligence chips to Meta Platforms, deepening strategic ties between two major AI infrastructure players and intensifying competition with market leader Nvidia.

Under the five-year agreement with AMD, the chipmaker will provide up to six gigawatts of AI compute capacity, beginning with one gigawatt of its forthcoming MI450 flagship GPUs. Initial shipments are scheduled to begin in the second half of 2026. The deployment will run on AMD's Helios rack-scale architecture, co-developed with Meta through the Open Compute Project, integrating custom MI450-based Instinct GPUs with 6th Gen EPYC CPUs, codenamed "Venice," powered by ROCm software.

The agreement significantly expands the companies' existing partnership. Meta has already deployed millions of EPYC CPUs and large clusters of Instinct MI300- and MI350-series GPUs across its global infrastructure. The new collaboration aligns silicon, systems and software roadmaps to optimize performance-per-dollar-per-watt for Meta's AI workloads, particularly inference — the computing process that enables AI systems to generate responses in real time.

As part of the transaction, AMD issued Meta a performance-based warrant to acquire up to 160 million shares — roughly 10% of AMD's outstanding stock — at a nominal exercise price. The warrant vests in tranches tied to shipment milestones, beginning with the first gigawatt of Instinct GPU deliveries and scaling to six gigawatts. Vesting is also linked to AMD achieving escalating stock price targets, reportedly up to $600 per share, along with Meta meeting specific technical and commercial benchmarks.

The structure mirrors a similar agreement AMD struck last year with OpenAI, reinforcing AMD's emergence as a viable second source for hyperscalers seeking alternatives to Nvidia's dominant ecosystem.

The deal comes just one week after Meta committed to deploying millions of Nvidia chips across its AI data centers, including Nvidia's new standalone CPUs and next-generation Vera Rubin systems. The parallel agreements underscore Meta's strategy of diversifying compute suppliers while scaling AI capacity at an unprecedented pace.

Those deals are part of Meta's broader commitment to invest up to $600 billion in the United States by 2028 to build data centers and supporting infrastructure. The company plans to operate 30 data centers globally, including 26 in the U.S. Its two largest AI campuses are currently under construction: the 1-gigawatt Prometheus site in New Albany, Ohio, and the 5-gigawatt Hyperion facility in Richland Parish, Louisiana — projects that highlight the massive scale of Meta's AI ambitions.

Despite AMD's growing momentum, Nvidia continues to control roughly 90% of the AI accelerator market and remains the dominant force in advanced AI hardware. Still, securing Meta as a multi-gigawatt customer provides AMD with significant multi-year revenue visibility and a strategic foothold at the center of the global AI infrastructure buildout.

For Meta, the dual-supplier strategy reduces reliance on a single vendor while maintaining flexibility to continue sourcing from Nvidia, AMD and internal chip development efforts. The expanding web of supply agreements, equity warrants and long-term capital commitments illustrates how the AI arms race is reshaping relationships between chipmakers and hyperscalers — with billions of dollars and future platform dominance at stake.