Google is in deep talks with Meta Platforms to supply Meta with Google’s custom AI chips, according to a report from The Information. The news created an immediate reaction across markets as Nvidia traders watched their positions turn red.
According to the report, Meta is preparing to spend billions to use Google’s tensor processing units, or TPUs, inside its data centers starting from 2027. The same report claimed Meta may also start renting TPUs from Google Cloud next year. Alphabet, Google’s parent company, jumped 2.7% in late trading, while Nvidia dropped 2.7%, signaling how fast investors reacted to the idea of Meta shifting part of its AI hardware stack.
Google expands TPU deals as Meta examines hardware
The Information also claimed that one unnamed person familiar with the talks said Meta plans to test TPUs for long-term use. Meta is one of the world’s heaviest spenders on AI infrastructure, so any change in its chip choices matters far beyond the company itself. Google’s TPUs were developed more than ten years ago for its internal AI work. They are ASIC chips designed for a single job: running and training AI models.
While Nvidia’s GPUs were originally made for video game graphics and later adopted for AI because they handle huge amounts of data, TPUs were built only for AI tasks. This is why Meta is looking into testing them now, as the firm continues to look into options in a market where Nvidia still controls most of the demand. Google already signed a deal to supply up to 1 million TPUs to Anthropic.
Jay Goldberg, an analyst at Seaport, called the Anthropic deal “a really powerful validation” and added, “A lot of people were already thinking about it, and a lot more people are probably thinking about it now.” The Meta talks also increased the attention, especially since it spends more on data centers than most companies in the world.
Asian markets also reacted fast, with South Korean firm IsuPetasys, which supplies multilayered boards to Alphabet, surged 18% to an intraday record. Meanwhile, in Taiwan, MediaTek gained almost 5% as traders priced in potential demand growth linked to Google’s chip ecosystem. Investors in Asia were responding to the idea that Google may expand its hardware partnerships well beyond its own products.
TPUs were built into Google’s internal systems, including AI models developed by Google and DeepMind. Engineers working on Google’s Gemini models share what they learn with Google’s chip designers, which helps shape new versions of TPUs. That feedback loop also gives AI teams the ability to customize hardware for their own workloads, something that Meta is now studying.
