Google is significantly stepping up its efforts to compete with Nvidia in the artificial intelligence chip race, and Meta is emerging as a potential multi-billion dollar customer, The Information reported late Monday. For years, Google has restricted its custom tensor processing units (TPUs) to its own cloud data centers and leased them to companies running large-scale artificial intelligence workloads.
But The Information reports that Google is now marketing the chips to customers in the hope they can deploy them in their own data centers, marking a major shift in its strategy.
One of those customers is Meta Platforms Inc (NASDAQ:META).
The parent company of Facebook and Instagram is reportedly in talks to spend billions of dollars to integrate Google's TPUs in its data centers starting in 2027, while also planning to lease TPU capacity from Google Cloud as early as next year. Meta currently relies heavily on Nvidia’s GPUs to build its AI infrastructure.
Alphabet Inc (NASDAQ:GOOGL) shares rose 2.1% in after-hours trading after the news was announced, while NVIDIA Corporation (NASDAQ:NVDA) shares fell 1.8%.
If the deal is completed, it will be a strong confirmation of Google's ambitions in the hardware field. According to "The Information", Google has told potential customers, including high-frequency trading companies and large financial institutions, that deploying TPUs (terminal processing units) locally can help them meet strict security and compliance requirements for sensitive data.
The stakes are high. Executives within Google Cloud said that expanding the application range of TPUs could help the company capture up to 10% of Nvidia's annual revenue, which is worth billions of dollars.
As the demand for artificial intelligence computing explodes, Nvidia continues to dominate the supply chain, and Google's move to deploy TPUs directly into customer facilities signals that the artificial intelligence chip war has entered a more intense stage.