Meta Taps AMD for $100B AI Infrastructure
Meta is reportedly securing $100 billion worth of AI infrastructure from AMD, in what is being described as a landmark deal for computing power. The deal is centered on AMD's Instinct GPUs and EPYC CPUs for Meta's AI racks, with roadmaps aligned for Meta's workloads starting in the second half of 2026. This represents a significant investment to expand Meta's AI capabilities and compute capacity.
- This multi-year agreement commits Meta to acquiring up to 6 gigawatts of AI computing power from AMD, with the first 1-gigawatt deployment scheduled for the second half of 2026. The deal is part of a broader trend of large tech firms securing long-term hardware supply to power future AI systems. - The initial hardware will include a custom AMD Instinct GPU based on the MI450 architecture, paired with 6th generation EPYC CPUs, codenamed "Venice". This hardware will be built on the Helios rack-scale architecture, which was co-developed by Meta and AMD through the Open Compute Project. - As a key component of the deal, AMD has issued Meta a performance-based warrant for up to 160 million shares of its common stock. This warrant vests in stages as Meta achieves specific GPU shipment milestones and AMD's stock reaches certain price thresholds, potentially giving Meta a stake of up to 10% in the chipmaker. - This partnership is a significant part of Meta's strategy to diversify its AI hardware suppliers and reduce its reliance on Nvidia, which currently dominates the AI chip market with an estimated 80-90% market share. Meta has also recently signed a multi-year deal with Nvidia for millions of its Blackwell and upcoming Rubin generation GPUs. - The AMD deal is primarily aimed at supporting Meta's inference workloads, which is the operational phase of running AI models, while the more computationally intensive training tasks will still heavily rely on Nvidia's hardware. Inference is estimated to account for 80-90% of AI computing energy usage. - Meta's total capital expenditure for 2026 is projected to be between $115 billion and $135 billion, a significant increase from $72.2 billion in 2025, with a large portion dedicated to scaling AI infrastructure. The company has stated plans to spend at least $600 billion on U.S. data centers and related infrastructure by 2028. - The scale of this investment reflects the immense power requirements for modern AI. A single large hyperscale data center can consume over 100 megawatts of power, and AI-optimized server racks require 40-60+ kW, compared to 5-15 kW for traditional racks. - This agreement and a similar recent deal with OpenAI are helping to solidify AMD's position as a viable competitor to Nvidia in the AI accelerator market. AMD's AI accelerator market share is projected to grow from around 9% in 2025 to over 15% by the end of 2026.
