Meta is widening its AI hardware strategy with a large AMD agreement that could reshape how it sources the chips behind its data centers. The companies announced Tuesday that Meta plans to purchase potentially up to $100 billion worth of AMD chips, enough to drive roughly six gigawatts of data center power demand.
The deal is not only about buying more processors. It also links Meta's future AI infrastructure needs to AMD's performance through a stock warrant arrangement, while giving Meta another major supplier as it works toward what Mark Zuckerberg calls personal superintelligence.
A chip deal built around scale
Under the multiyear agreement, Meta will purchase AMD's MI540 series of GPUs and AMD's latest generation of CPUs. The size of the commitment is notable: the companies described a potential purchase value of up to $100 billion.
That scale fits with Meta's broader data center ambitions. The company has pledged to invest at least $600 billion in U.S. data centers and AI infrastructure over the next several years, including a projected capital expenditure spend of $135 billion in 2026.
Meta has also recently unveiled plans for a $10 billion gas-powered data center campus in Indiana designed for 1 gigawatt of compute capacity. The AMD agreement points to the same basic reality: Meta's AI roadmap depends on enormous amounts of computing infrastructure, and that infrastructure needs chips, power, and long-term supplier relationships.
Why CPUs matter in Meta's AI infrastructure
The agreement includes both GPUs and CPUs, which matters because CPUs are becoming a more important part of the AI inference compute stack. According to the source, CPUs are efficient, easier to scale, and help companies avoid being tied solely to Nvidia.
AMD CEO Lisa Su framed the opportunity directly during an investor briefing Tuesday morning, saying, "The CPU market is absolutely on fire." She said demand has continued to grow as AI infrastructure deployments expand, including inferencing and agentic AI.
For Meta, that makes the AMD deal more than a supply purchase. It gives the company another route for scaling AI systems while building a compute base that is not dependent on a single vendor. The source article says AMD has been slowly gaining ground as AI firms seek to reduce reliance on Nvidia, which has been the longstanding leader in AI chips and has charged a premium for that position.
The equity structure gives AMD a second upside
As part of the agreement, AMD issued Meta a performance-based warrant for up to 160 million shares of AMD common stock. That would amount to about 10% of the company. The exercise price is $0.01 per share, and the warrant is structured to vest alongside certain milestones.
The full stock award is conditional on AMD's share price. Per The Wall Street Journal, AMD's share price would need to hit $600 for Meta to receive its final tranche. AMD's stock closed at $196.60 on Monday.
This structure connects the commercial chip purchase to AMD's market performance. If AMD reaches the required milestones and share price conditions, Meta could receive a significant equity position. If those conditions are not met, the full award would not vest.
The arrangement also echoes another recent AMD partnership. Last October, AMD and OpenAI struck a similar deal that traded equity for an agreement to buy chips.
Meta is diversifying, not choosing one supplier
The AMD partnership arrives a couple of weeks after Meta struck a multiyear deal to expand its data centers with millions of Nvidia's latest CPUs and GPUs. That timing is important because it shows Meta adding AMD capacity while still working with Nvidia at a large scale.
Meta is also working on its own in-house chips, though it has reportedly hit delays. Taken together, the source points to a three-part compute strategy:
- Buy from Nvidia for large-scale data center expansion.
- Buy from AMD through a multiyear agreement covering GPUs and CPUs.
- Develop in-house chips while navigating reported delays.
That mix reflects the pressure created by AI infrastructure demand. Meta needs enough compute to support its AI plans, but it also needs flexibility across suppliers, chip types, and data center projects.
How the deal fits personal superintelligence
Mark Zuckerberg said Meta's partnership with AMD is "an important step" as the company diversifies its compute and works toward "personal superintelligence." Zuckerberg has defined personal superintelligence as AI systems designed to deeply understand and empower individuals in their everyday lives.
The AMD deal does not, by itself, explain what those systems will look like. What it does show is the infrastructure direction behind the ambition. Meta is tying its AI future to massive data center investment, large chip purchase agreements, and a broader supplier base.
For AMD, the agreement strengthens its position in the AI infrastructure market at a time when companies are looking beyond a single dominant chip provider. For Meta, it adds another major path toward the compute capacity needed for its next generation of AI systems.