Meta Platforms has struck a multi-year deal worth several billion dollars with Google to lease artificial intelligence chips for training its next wave of AI models.
At the centre of the agreement are Google’s tensor processing units, or TPUs — chips the company has designed specifically for machine learning tasks. Google has been pushing these processors as an alternative to the graphics chips made by NVIDIA, which remain the dominant force in AI training hardware.
Growing Race for AI Computing Power
The move reflects how seriously large technology firms are taking the need for computing power. As AI systems grow more complex, companies are racing to secure enough capacity to keep their models competitive. Meta, rather than relying on a single supplier, appears to be spreading its investments across multiple chipmakers.
The social media giant has already committed heavily to AI infrastructure. Earlier this week, Advanced Micro Devices said it expects to sell as much as $60 billion worth of AI chips to Meta. The company also continues to work with NVIDIA for both current and upcoming processors, signalling a long-term buildout strategy.
For Google, the partnership provides an opportunity to expand the reach of its in-house hardware. By leasing TPUs externally, the company can strengthen its cloud business while making better use of its custom silicon investments.
There are also reports that Meta is exploring the possibility of purchasing TPUs outright for use in its own data centres as early as next year, though details remain unclear. Separately, Google is said to be working with a major investment firm to set up a joint venture that would lease TPUs to additional customers, potentially widening access while sharing financial exposure.
The agreement underscores a broader shift in the technology sector. Even as Meta and Google compete in advertising, social media and AI services, they are increasingly partnering when it comes to infrastructure. In the AI era, access to powerful chips has become one of the most important assets shaping both competition and collaboration in Silicon Valley.
Also Read: Meta’s new AI team delivered its first key models internally this month, CTO says
