Google & Meta Team Up on AI Chips

This partnership will see Meta lease Google's custom Tensor Processing Units (TPUs), specialized chips designed for AI workloads. The multi-year, multi-billion dollar agreement provides Meta with access to Google's advanced hardware to train and deploy its next-generation AI models. For Google, this deal represents a significant step in monetizing its custom hardware and bolstering its cloud AI services. The move is part of a broader strategy by Meta to diversify its sources for high-performance chips and mitigate its dependence on a single supplier. This follows reports of Meta halting its own advanced in-house AI chip development project, codenamed "Olympus," due to its complexity and the risk of falling behind competitors like Google and OpenAI. The company had also been developing a simpler chip called "Iris," which also faced setbacks. Beyond the Google deal, Meta has been actively securing its AI future on multiple fronts. The company recently announced a deal with AMD for AI chips and remains one of Nvidia's largest customers for its high-end processors. This multi-supplier approach is crucial as Meta plans to invest heavily in AI infrastructure. This collaboration intensifies the competition in an AI hardware market overwhelmingly dominated by Nvidia, which holds a market share estimated to be as high as 95% in AI data center chips. Nvidia recently reported a 75% year-over-year growth in its data center business, underscoring the massive demand for AI computing power. The market for AI hardware is expanding rapidly, with the global market valued at nearly $87 billion in 2024 and projected to grow significantly. This growth is fueled by the escalating computational needs of increasingly complex AI models across all major tech companies.