Amd and amd announce 6-gigawatt GPU deal as part of AI build-out
Meta and amd are entering a multiyear agreement in which Meta will purchase upward of 6 gigawatts' worth of AI chips as part of its massive AI build-out, a deal announced Tuesday. The agreement includes a 160 million share issuance from AMD to Meta that vests in tranches tied to milestones, with the first tranche linked to shipment of the initial 1 gigawatt of chips.
Amd stock reaction
After the announcement, amd shares jumped as much as 10% in premarket trading before paring those gains in the regular session. AMD CFO Jean Hu said, "We expect this partnership to drive substantial multi-year revenue growth and be accretive to our non-GAAP earnings per share, marking another significant step forward in delivering on our ambitious long-term financial model. " Hu added, "The performance-based structure also tightly aligns AMD and Meta around execution and long-term value creation. "
Deal terms and stock grant
The multiyear pact calls for Meta to buy upward of 6 gigawatts of AI chips. AMD will issue Meta 160 million shares of common stock that will vest in a series of tranches if AMD hits certain milestones, the first of which will hit when AMD ships its first 1 gigawatt of chips.
GPUs, racks and timing
The first GPUs tied to the deal are AMD’s MI450 line, which will be deployed in the company’s Helios rack-scale data center systems, complete with EPYC CPUs, in the second half of the year. The timeline places initial hardware deliveries within the second half of the year as Meta begins integrating the MI450 GPUs into its Helios systems.
CPU buys and inference focus
Meta says it will also purchase a large additional CPUs, including AMD’s Venice chip and its next-generation Verano processor. The deal highlights how CPUs are becoming an increasingly important component for AI data centers as companies look toward running agentic AI services and focus on inference, or running models.
Nvidia agreement and chip pairs
Last week Meta announced a separate multiyear agreement with AMD rival Nvidia that will see that chipmaker provide Meta with millions of its Blackwell and Rubin GPUs. Nvidia also said Meta will host the first large-scale deployment of its Grace CPU servers. Nvidia usually pairs its Grace CPU with two of its Blackwell or Blackwell Ultra processors to form its GB200 and GB300 superchips that slot into its rack-scale platforms.
Spending plans and market reaction
Meta’s agreement with AMD comes as the company is set to spend upwards of $135 billion in capital expenditures throughout 2026 as part of its AI expansion plans, money that will pay for everything from data center construction and chips to training models. Meta, Amazon, Google, and Microsoft plan to collectively spend some $650 billion on AI this year.
That scale of investment has given some investors pause as they digest whether the investments will bear fruit. Meta has fared the best among the group of big spenders, with shares down just 2. 5% since it reported earnings on Jan. 28 and announced its investing plans. Google stock has dropped 8. 7% since its own news, while Amazon has fallen 11. 9% and Microsoft has declined 15. 5%.
Chip stocks more broadly have cooled as Wall Street weighs questions of an AI bubble and worries that custom chips built by Amazon, Google, Meta, and Microsoft could pose a threat to GPU supremacy in the data center. In November, Meta was in talks with Google to use its TPU chips for AI applications. Analysts say it will be difficult for those chips to supplant the more general-purpose AI processors Nvidia and AMD build.
Daniel Howley is named in the context of the original coverage.
Closing: The agreement ties a large, milestone-driven stock grant to hardware shipments, sets MI450 GPUs for Helios deployment in the second half of the year, commits Meta to large CPU purchases including Venice and Verano, and comes amid sweeping AI spending plans and separate GPU deals with Nvidia.