A multi-year, multi-generation agreement to install six gigawatts of AMD Instinct GPUs for Meta’s AI infrastructure was announced by Meta Platforms and AMD on Tuesday. The deal could bring in tens of billions of dollars for the chipmaker and change the competitive environment in the AI hardware market.
The collaboration is Meta’s most direct attempt to date to expand its supply chain for AI computation beyond Nvidia, which only a week ago signed its own comprehensive agreement with Meta. AMD granted Meta a performance-based warrant under the deal for up to 160 million shares of AMD common stock, or around 10% of the business. The warrant is set to vest when shipping milestones and stock price benchmarks are met.
Timeline And Structure Of The Deal
Powered by a custom AMD Instinct GPU based on the MI450 architecture and 6th-generation AMD EPYC CPUs codenamed “Venice,” running ROCm software and built on the AMD Helios rack-scale architecture, shipments supporting the first gigawatt of deployment are expected to start in the second half of 2026. AMD and Meta collaborated on the Helios system as part of the Open Compute Project.
As Meta pushes the limits of AI at a never-before-seen scale, we are thrilled to deepen our strategic cooperation with them,” AMD CEO Lisa Su said in a statement. The deal is “an important step for Meta as we diversify our compute,” according to Meta CEO Mark Zuckerberg, who also stated that AMD “will be an important partner for many years to come.”
The relationship is anticipated to “drive substantial multi-year revenue growth and be accretive to our non-GAAP earnings per share,” according to Jean Hu, CFO of AMD.
Context Of Strategy
The arrangement closely resembles a historic one AMD made with OpenAI in October 2025 for 6 gigawatts of Instinct GPUs, which also had the same warrant structure of 160 million shares. That OpenAI transaction was credited with helping AMD establish itself as a viable rival to Nvidia in the hyperscale AI sector, with Reuters reporting that it might generate over $100 billion in revenue for AMD over four years from cascading demand.
Only a few days after strengthening its ties with Nvidia with a different multi-year agreement involving millions of Blackwell and next-generation Vera Rubin GPUs, Meta decided to sign with AMD. The two agreements collectively highlight Meta’s approach of obtaining computing power from several providers in pursuit of what Zuckerberg has called the company’s objective of providing “personal superintelligence” to consumers everywhere. According to Meta, it intends to spend up to $135 billion on AI infrastructure by 2026.
Implications For Competition
AMD’s data center AI plan has been validated with the addition of Meta, one of the biggest spenders on AI infrastructure globally, as an anchor customer for its MI450 platform. According to earlier estimates by analysts, the market for chips might reach $20 billion for every gigawatt of AMD AI capacity. Tuesday’s pre-market session saw a decline in AMD shares, and early trading saw a decline in Meta as well.
With Meta positioned as a primary customer for next EPYC processors, including “Verano,” a next-generation CPU with workload-specific improvements, the agreement places AMD’s Helios rack-scale systems at the core of one of the biggest AI buildouts in the industry.