Meta is in discussions to purchase billions of dollars worth of artificial intelligence chips from Google, according to a report published by The Information. The talks come as Meta accelerates development of its Llama AI models and seeks expanded compute power to support large-scale training and deployment.
Sources cited in the report said Meta is evaluating Google’s custom Tensor Processing Units as an alternative to NVIDIA hardware, which remains in extremely high demand. Google’s TPUs, used internally to power Search, Gmail, and its own AI models, have evolved into a major competitive asset as cloud providers race to supply AI infrastructure.
If finalized, the agreement would represent a rare collaboration between two of the world’s largest technology firms, both of which are competing aggressively in generative AI. Analysts say the deal would also highlight the growing pressure on major companies to secure reliable semiconductor capacity as global demand outpaces supply.
Meta has been scaling its AI investments across platforms including Facebook, Instagram, WhatsApp, and its metaverse division. The company’s growing need for high-performance compute has pushed it to explore multiple chip sources, including custom silicon development and external partnerships.
Google, meanwhile, is positioning its TPU lineup as a cost-effective, high-efficiency alternative for companies running large model workloads. A supply partnership with Meta would further strengthen Google Cloud’s standing in the enterprise AI market.
While no final decision has been announced, industry observers say the outcome of the talks could influence broader competition in AI hardware, cloud revenue dynamics, and long-term strategic alliances across Big Tech.