Meta has signed a long term agreement with AMD to strengthen its artificial intelligence infrastructure, just days after confirming a large scale deployment of Nvidia chips. The company is clearly spreading its bets as demand for AI computing power continues to grow across its platforms.
Up To 6 Gigawatts of AI GPUs for Data Centers
The partnership will see Meta deploy as much as 6 gigawatts of AMD graphics processing units inside its AI data centers. These systems will also use AI focused CPUs. Early deliveries of AMD’s MI450 GPUs, built into Helios rack scale servers, are expected to begin later this year.
This move supports Meta’s broader plan to build advanced computing capacity for training and running large AI models across its services.
Performance Based Share Warrant for Meta
As part of the agreement, AMD has issued a performance linked warrant that allows Meta to acquire 160 million shares, close to ten percent of the chipmaker. The shares will vest in phases as Meta purchases more GPU capacity, starting with the first one gigawatt deployment and continuing until the full six gigawatts are installed.
The structure also depends on AMD’s stock performance and key technical and commercial milestones, aligning the interests of both companies over the long term.
Also Read: CodeBrain 1 Agent - Is This The Smartest AI Coding Tool Yet?
Massive AI Spending Plan
Meta recently said it could spend up to 135 billion dollars this year on capital expenditure, mainly to expand its AI infrastructure. The company is planning around 30 data centers, with most of them located in the United States.
This investment is aimed at keeping pace with global AI leaders and meeting the growing computing needs of its models and services.
A Big Opportunity for AMD
For AMD, the deal marks an important step in challenging Nvidia’s strong position in the AI chip market. Nvidia currently controls the majority share and has seen huge demand for its Grace Blackwell systems since their launch.
Helios servers are AMD’s first large scale attempt to compete directly in this segment. Analysts believe the Meta agreement could be worth tens of billions of dollars over several years because deploying six gigawatts of compute capacity will take time.
Custom Hardware Gives AMD an Edge
One notable detail is that the first phase of deployment will use customized GPUs designed for Meta’s requirements. This level of tailoring has not been publicly linked to Nvidia’s current agreement with the company and could help AMD secure more hyperscale customers.
Meta’s Multi Vendor Approach
Meta has been a long time customer of both AMD and Nvidia and also develops its own chips. Reports have suggested that the company has explored using Google’s tensor processing units in the future.
By working with multiple partners, Meta is building a flexible AI hardware stack that reduces dependency on a single supplier while ensuring enough computing power for the next phase of AI growth.
Also Read: ChatGPT Image Generator - 12 Essential Tricks In 2026