Meta has reached a major new supply agreement with AMD, committing to purchase 6 gigawatts’ worth of AMD’s artificial intelligence chips as part of a multiyear effort to expand its computing infrastructure. The deal illustrates Meta’s strategy to diversify beyond a single chip supplier while continuing to scale the hardware needed for its artificial intelligence roadmap. Company executives positioned the arrangement as a way to support large scale inference workloads and long term ambitions around more powerful consumer facing systems.
Under the agreement, Meta plans to deploy custom versions of AMD’s Instinct GPUs, with shipments expected to start in the second half of 2026, according to the companies. AMD’s stock was up by as much as nearly 12% premarket following the announcement, reflecting investor optimism about AMD’s growing role in the artificial intelligence chip market. The deal would also allow Meta to purchase up to 10% of AMD’s stock, with the shares vesting based on specific shipping milestones that the two companies have agreed upon, further aligning their incentives over the life of the partnership.
Meta recently deepened its relationship with Nvidia through a “multigenerational” agreement that commits the company to use millions of Nvidia’s current and next generation chips for its artificial intelligence buildout, while AMD has separately signed a multiyear partnership to provide its artificial intelligence chips to OpenAI. Meta chief executive Mark Zuckerberg described the AMD collaboration as a key step in diversifying the chips used to power Meta’s artificial intelligence initiatives and in deploying efficient inference compute to deliver what he called personal superintelligence. Together, the deals highlight how leading technology companies expect that vast amounts of infrastructure will be necessary to meet rising artificial intelligence demand, with 1 gigawatt roughly equivalent to the power needed to run the homes of a midsize city like San Francisco.
