Nvidia Corp.’s stock fell after a report said Meta Platforms Inc. is in talks to use Google’s AI chips, suggesting Google is challenging Nvidia’s dominance in AI hardware.
Meta may start using Google’s tensor processing units (TPUs) in its data centers by 2027, and could rent chips from Google Cloud as soon as next year, according to The Information, citing an anonymous source familiar with the discussions.
Shares of Alphabet, Google’s parent company, rose as much as 2.7% in late trading, while Nvidia dropped 2.7% at one point.
Nvidia’s graphics processing units (GPUs) are the industry standard for AI computing, used by companies from Meta to OpenAI. TPUs, specialized chips designed for AI and machine learning, could become a viable alternative.
Google already agreed to supply up to 1 million TPUs to AI startup Anthropic. Analysts say a deal with Meta would further validate TPUs as a competitor to Nvidia. “A lot of people were already thinking about it, and a lot more people are probably thinking about it now,” said Seaport analyst Jay Goldberg.
According to Bloomberg Intelligence analysts Mandeep Singh and Robert Biggar, Meta’s likely TPU adoption could accelerate Google Cloud demand. They estimate that Meta’s 2026 capital spending could include $40-$50 billion on AI chip capacity.
Asian stocks tied to Alphabet also surged. In South Korea, IsuPetasys Co., which supplies components for Google, jumped 18%. Taiwan’s MediaTek Inc. rose nearly 5%.
TPUs were created over a decade ago for AI tasks and have gained attention outside Google as a way to train and run complex AI models. GPUs, initially made for graphics rendering, are widely used for AI but companies are seeking alternatives to reduce reliance on Nvidia.
Google’s TPUs benefit from lessons learned from its AI teams, including DeepMind, and are tailored specifically for AI workloads. Analysts say their efficiency and power will determine whether they can compete long-term with Nvidia’s GPUs.