Meta and AMD Ink AI Chip Deal Worth Up to $100B
Meta has signed a long-term agreement with chipmaker AMD for a deal potentially worth up to $100 billion. The partnership gives Meta privileged access to next-generation AI computer infrastructure and includes an option for Meta to acquire a 10% stake in AMD. The move signals a strategic bet on building a proprietary, AI-native business infrastructure to power future platforms and services, according to CNET.
- This deal is part of Meta's broader strategy to diversify its AI chip suppliers beyond market leader Nvidia, on which it spent an estimated $10 billion in 2023 alone. Just a week prior, Meta announced a separate multi-year agreement to purchase millions of AI chips from Nvidia, including its upcoming Vera Rubin platform. - The agreement's structure is a "chips for shares" arrangement where AMD issued Meta a performance-based warrant to acquire up to 160 million shares of AMD stock, roughly a 10% stake, at an exercise price of just $0.01 per share. Vesting is tied to hitting specific GPU shipment milestones and AMD's stock price reaching as high as $600. - AMD will supply Meta with up to 6 gigawatts of computing capacity, starting with its forthcoming MI450 hardware in the second half of 2026. Each gigawatt of computing power is valued in the "double-digit billions." - This partnership mirrors a similar deal AMD struck with OpenAI in October 2025, which also included a 6-gigawatt chip supply and an option for a 10% stake, signaling a pattern of deep integration with major AI players. - The deal is a component of Meta's massive capital expenditure plan for 2026, which is projected to be between $115 billion and $135 billion, largely driven by AI infrastructure investments. This level of spending is part of a larger trend, with Big Tech companies expected to collectively spend nearly $650 billion on AI assets in 2026. - While Nvidia currently holds a dominant market share of 80-95% in AI GPUs, this deal significantly strengthens AMD's position as the primary alternative. AMD aims to capture a double-digit share of the AI accelerator market, which is projected to grow toward $1 trillion by 2030. - Meta has also been developing its own custom silicon, the Meta Training and Inference Accelerator (MTIA), since 2020 to handle its specific AI workloads, particularly for recommendation models on Facebook and Instagram. This three-pronged strategy involves using chips from Nvidia, AMD, and its own in-house designs to avoid vendor lock-in and optimize for different tasks.
