Nvidia shares fall after a possible Meta and Google partnership

Nvidia Corp. shares fell after a report said Meta Platforms Inc. is in talks to spend billions of dollars on Google’s artificial intelligence chips. The move signals new traction for Google as it works to build a competitor to Nvidia’s dominant AI accelerators.

The Information reported that Meta is discussing the use of Google’s tensor processing units, or TPUs, in its data centers in 2027. The outlet, citing a person familiar with the talks, also said Meta may rent chips from Google’s cloud division next year.

Alphabet Inc. shares rose as much as 2.7% in late trading following the report. Nvidia dropped as much as 2.7%.

A deal would strengthen TPUs as an alternative to Nvidia’s chips, which remain the preferred hardware for companies developing and operating artificial intelligence systems. Google has already agreed to supply as many as 1 million TPUs to Anthropic PBC, though Nvidia continues to hold the leading position in the market.

After that earlier deal, Seaport analyst Jay Goldberg described the arrangement as strong validation for Google’s chip strategy. He said it demonstrated increasing interest in TPUs among companies evaluating options for AI development.

Meta declined to comment on the latest report. Google did not respond to requests for comment.

Analysts at Bloomberg Intelligence said Meta’s potential use of Google’s chips indicates that large model developers may turn to Google as a secondary supplier of accelerators for inferencing. They estimated Meta will direct $40 billion to $50 billion toward inferencing-chip capacity next year as part of planned capital expenditures of at least $100 billion in 2026. They said Google Cloud could see faster consumption growth than other major cloud providers as demand builds for TPUs and Gemini models.

Asian stocks linked to Alphabet rose in early Tuesday trading. IsuPetasys Co. in South Korea gained 18% to a record, while MediaTek Inc. in Taiwan advanced almost 5%.

Google introduced TPUs more than a decade ago for artificial intelligence tasks and later adapted them as accelerators for machine learning across its products. The company’s DeepMind and Gemini teams have influenced the chips’ design as Google integrates insights from model development into its hardware efforts.