Google & Meta Team Up on AI Chips
Google has signed a multibillion-dollar deal to supply its custom AI chips to Meta. The alliance is a direct challenge to Nvidia's market dominance, as two of the biggest AI players team up to reduce their reliance on a single supplier. It's a major escalation in the "AI chip wars" as hyperscalers look to control their own hardware destiny.
The multi-year, multi-billion dollar agreement will see Meta rent Google's custom Tensor Processing Units (TPUs) to train and run its next-generation AI models. This is a significant move for Google, which has historically used its TPUs almost exclusively for its own products but is now aiming to capture up to 10% of Nvidia's annual revenue by selling to third parties. This partnership is part of Meta's broader "multipronged silicon strategy" to diversify its AI chip suppliers. The company recently announced a deal to buy up to $60 billion worth of AI chips from AMD and remains one of Nvidia's largest customers, with plans to purchase millions of their processors. Nvidia currently dominates the AI data center chip market with a staggering 95% market share. The high demand, however, has led to tight supply and sky-high prices, pushing major AI players like Meta to seek out alternatives and reduce their dependence on a single supplier. While Meta is renting Google's hardware, it is also developing its own custom silicon. The company's "Meta Training and Inference Accelerator" (MTIA) chips are designed to handle its specific workloads, particularly for AI-powered recommendation systems on Facebook and Instagram. The deal also includes discussions for Meta to potentially purchase TPUs directly from Google for its own data centers as early as next year. This second phase would further solidify Google's position as a major player in the external AI infrastructure market. For Google, this represents a major commercialization of its long-standing TPU project. The company has already signed a massive deal with AI startup Anthropic, valued at tens of billions of dollars, for access to its TPUs, signaling an aggressive new phase in the AI chip wars.
