Facebook owner Meta has locked in one of the largest AI hardware commitments in the industry, signing a multibillion‑dollar deal with Advanced Micro Devices (AMD) to secure next‑generation AI chips for its data centers.
The multi‑year agreement will see AMD supply up to 6-gigawatts of AI GPU capacity, positioning Meta to scale its AI infrastructure well beyond its current deployments.
Estimated at $60 billion, the deal covers multiple generations of AMD chips, aligns both companies’ hardware and software development roadmaps, and includes a significant equity arrangement that could give Meta a roughly 10% stake in AMD over time.
What the Deal Actually Covers
The agreement goes beyond a straightforward chip purchase. The first deployment will use a custom AMD Instinct GPU based on the MI450 architecture, designed and optimized specifically for Meta’s workloads, with shipments expected to begin in the second half of 2026.
Meta will also purchase AMD’s latest generation of EPYC CPUs as part of the deal, and the infrastructure will run on AMD’s Helios rack-scale architecture, which represents the first large-scale competition to Nvidia’s Grace Blackwell systems.
Meta’s infrastructure head Santosh Janardhan was direct about the company’s multi-vendor approach saying, “We don’t believe that a single silicon solution will work for all of our workloads. There’s a place for Nvidia, there’s a place for AMD, and there’s a place for our own custom silicon as well. We need all three.”
And beyond the hardware agreement, AMD issued Meta a performance-based warrant allowing it to buy up to 160 million shares of AMD common stock, structured to vest as specific milestones associated with Instinct GPU shipments are achieved. At current valuations, hitting all those milestones would give Meta approximately a 10% stake in AMD. Shares in AMD, which has a market capitalisation of over $300 billion, surged more than 6% after the announcement.
This structure closely mirrors a deal AMD signed with OpenAI in October 2025, where the ChatGPT maker was similarly offered an equity stake tied to chip purchases.
Where This Fits in Meta’s Broader AI Spending
Meta has pledged to invest at least $600 billion in U.S. data centers and AI infrastructure by 2028, with projected capital expenditure of $135 billion in 2026 alone, up from $72 billion in 2025.
The AMD deal comes days after Meta announced a separate multiyear agreement with Nvidia to deploy millions of its latest CPUs and GPUs. Meta is also developing its own in-house chips, although those efforts have reportedly hit delays.
For Meta, the deal is about securing reliable AI infrastructure supply at massive scale, especially as it rolls out generative AI (GenAI) across its apps. As such, locking in 6 gigawatts of compute gives the company a defined roadmap for expanding its AI infrastructure over the next several years.
For AMD, the deal provides a guaranteed revenue base across multiple chip generations, and a high-profile customer that helps make AMD’s case to other buyers that its hardware can compete at hyperscaler scale.
However, there is a question of how the circular financing structure plays a role here, as it’s an arrangement where one company invests in another that then uses that capital to buy its products, ultimately binding both parties’ financial outcomes to each other.
Now whether this Meta-AMD deal reflects a genuine market dynamic or a financial structure that overstates real demand is a question that remains unanswered.
