Google and Meta Team Up on AI Chips
Google has signed a multibillion-dollar deal to supply AI chips to Meta, escalating the hardware rivalry with Nvidia. The move is a significant push to diversify the global AI hardware ecosystem beyond Nvidia's dominance. This partnership is expected to accelerate the development of next-gen AI platforms for both companies.
The deal centers on Google's Tensor Processing Units (TPUs), custom-built chips designed to accelerate AI workloads. Meta is initially renting the TPUs in a multi-year, multi-billion dollar agreement, a move that helps Google justify its massive investments in custom silicon and cloud infrastructure. There are also discussions about Meta purchasing the TPUs directly for its own data centers as early as 2027. This partnership is part of a broader strategy by Meta to diversify its hardware supply chain and reduce its heavy reliance on Nvidia. Nvidia currently dominates the AI chip market, with some estimates putting its market share between 80% and 90%. This dominance has led to high prices and supply bottlenecks, prompting major AI players to seek alternatives. Meta's decision to partner with Google also follows reported technical challenges and a scaling back of its own in-house chip development program, known as the "Meta Training and Inference Accelerator" or MTIA. The company had ambitious plans for custom chips codenamed "Olympus" and "Iris" but faced difficulties, making external partnerships more critical. This agreement is not exclusive; Meta is aggressively securing computing power from multiple sources. The company recently announced a massive deal with AMD, potentially worth up to $60 billion, for its AI chips. Meta also continues to be one of Nvidia's largest customers, with plans to operate over 1.3 million GPUs by the end of 2025.
