Google and Meta Team Up on AI Chips
Google and Meta have signed a multibillion-dollar deal for AI chips, a direct challenge to Nvidia's market dominance. The pact signals an intensifying race among tech giants to control their own AI hardware supply chains and reduce reliance on a single vendor for critical infrastructure.
This multi-billion dollar, multi-year deal will see Meta lease Google's custom-built Tensor Processing Units (TPUs) to power the training and inference of its next-generation AI models. The partnership could expand as early as 2027, with Meta potentially purchasing TPUs directly to install in its own data centers. The move comes as Meta faces setbacks in its own AI chip development. The company recently scrapped its most advanced in-house training processor, codenamed "Olympus," due to design complexities and concerns it couldn't compete with Nvidia's offerings. This has led Meta to also sign significant chip supply deals with both Nvidia and AMD, with the latter valued at over $100 billion. For Google, this partnership marks a significant step in commercializing its TPUs, which were initially developed for internal use. By supplying a major rival, Google is positioning its TPUs as a direct competitor to Nvidia's GPUs and aiming to capture a larger share of the booming AI hardware market. Nvidia currently holds a dominant position, controlling over 80% of the market for GPUs used in AI. The immense demand for AI computing has led to tight supply and high prices for Nvidia's hardware, prompting major tech companies to seek out alternatives and diversify their supply chains. News of the Google-Meta deal caused a dip in Nvidia's stock price, highlighting investor concerns about rising competition.
