Meta and AMD have entered into a significant partnership aimed at enhancing artificial intelligence infrastructure and accelerating the deployment of advanced AI models. The agreement, which spans multiple years and generations, involves a massive commitment to deploy up to 6 gigawatts of AMD Instinct GPUs to support Meta’s evolving tech landscape.
The foundation of this collaboration is the AMD Helios rack-scale architecture, which was introduced at the recent Open Compute Project Global Summit. Shipments supporting the first gigawatt deployment are expected to commence in the latter half of 2026. At the core of this initiative will be a specialized AMD Instinct GPU based on the MI450 architecture, specifically optimized for Meta’s unique workload requirements.
Dr. Lisa Su, chair and CEO of AMD, highlighted the significance of this extension of their strategic partnership with Meta. Emphasizing the urgency and scale of the AI advancements in play, she stated that this collaboration would ensure the provision of high-performance, energy-efficient infrastructure designed to support one of the largest AI deployments in the industry. Dr. Su remarked, “This multi-year, multi-generation collaboration aligns our roadmaps to deliver… accelerated one of the industry’s largest AI deployments.”
Mark Zuckerberg, founder and CEO of Meta, echoed similar sentiments, expressing enthusiasm for the long-term partnership. He remarked that this collaboration marks a critical step in diversifying Meta’s computational capabilities, anticipating that AMD will play a pivotal role in their technological landscape for years to come.
In tandem with the advancements surrounding GPUs, AMD and Meta are also expanding their collaboration on AMD EPYC processors. Meta has made extensive use of AMD EPYC CPUs over several generations, having integrated millions of these chips into their global infrastructure. As the demands of AI infrastructure become increasingly complex, CPUs remain a critical component in optimizing performance and efficiency.
As part of the broader agreement, AMD has issued Meta a performance-based warrant for up to 160 million shares of AMD common stock, contingent on the achievement of specific milestones tied to the deployment of the Instinct GPUs. This unique structure aims to align both organizations’ strategic interests, promoting a collaborative approach to reaching their mutual goals.
AMD’s EVP and CFO, Jean Hu, stated that the partnership is expected to yield substantial multi-year revenue growth, enhancing AMD’s non-GAAP earnings per share. The performance-based mechanism not only underscores mutual accountability but also emphasizes the long-term creation of value for both companies.
As AMD and Meta work together, their joint efforts are likely to set new standards for AI infrastructure, facilitating innovation and delivering AI-driven services to millions globally. This collaboration positions both companies to leverage their strengths, ultimately aiming to reshape the future of intelligent computing.
