Advanced Micro Devices and Meta Platforms announced an expanded strategic partnership on February 24, 2026 under which AMD will supply up to $60 billion worth of AI chips to Meta over five years, covering as much as 6 gigawatts of Instinct GPU capacity. As part of the arrangement, AMD granted Meta a performance-based warrant that could give the social media giant up to 10 percent of AMD’s equity, an unusual structure that ties a chip supplier’s ownership directly to how much its biggest customer ends up buying.
A custom chip built specifically for Meta
At the center of the deal is a custom version of AMD’s Instinct MI450 GPU, co-engineered with Meta and optimized specifically for inference workloads, the computational task of running trained AI models rather than training them from scratch. The agreement also includes two generations of AMD’s Epyc server CPUs. Shipments supporting the first gigawatt of deployment are scheduled to begin in the second half of 2026, according to AMD’s announcement, with Meta having directly contributed to the MI450’s design specifications.
Why the equity warrant is the real story
The performance-based warrant is what sets this deal apart from a typical chip supply contract. Rather than negotiating purely on price, AMD effectively offered Meta a stake in its own future success as an incentive to commit to large-scale purchases, a structure that aligns Meta’s incentives with AMD’s stock performance and signals just how badly AMD wants to prove it can be a credible alternative to Nvidia at scale. AMD shares rose more than 10 percent in premarket trading when the deal was announced, reflecting investor enthusiasm for the company’s growing foothold in a market Nvidia has dominated.
Meta’s motivation: diversifying away from Nvidia
For Meta, the deal is part of a broader strategy to avoid depending entirely on any single chip supplier. The AMD agreement runs alongside Meta’s separate custom silicon partnership with Broadcom for its in-house MTIA chips, giving the company three distinct sources of AI compute: Nvidia GPUs, AMD GPUs, and its own custom accelerators. Executives have argued that this diversification gives Meta stronger negotiating leverage on price and reduces the risk of supply bottlenecks if any one vendor faces production constraints.
What it means for AMD’s competitive position
The deal is widely viewed as a milestone in AMD’s attempt to break into a GPU market Nvidia has controlled for years. Landing a customer as large as Meta, willing to commit to $60 billion in purchases over five years, gives AMD a credible reference customer it can point to when pursuing other hyperscalers. Some analysts have framed the arrangement as evidence that Nvidia’s near-monopoly on AI training and inference chips is beginning to erode, while others caution that Nvidia’s software ecosystem, built around its CUDA platform, still gives it an advantage that a hardware deal alone does not immediately overcome.
The risk on both sides
The arrangement is not without risk for either company. AMD is betting a meaningful slice of its future equity value on Meta’s continued willingness to scale purchases as planned, and any slowdown in Meta’s AI capital spending could leave AMD short of the volumes needed to justify the warrant’s cost. Meta, meanwhile, is wagering that custom MI450 chips will perform well enough at scale to justify diverting purchasing away from Nvidia, in a market where software compatibility and developer familiarity still tend to favor incumbents.
What’s next
The first wave of MI450 shipments later this year will be the initial real-world test of whether the deal’s ambitious targets hold up. If AMD delivers chips on schedule and Meta’s inference workloads run efficiently on the new hardware, the arrangement could become a template other hyperscalers use to extract equity-linked concessions from chipmakers eager for guaranteed volume. If it falters, it may reinforce the view that unseating Nvidia’s ecosystem advantage takes more than favorable deal terms alone.
Photo: MDGovpics / BY via flickr