Meta Platforms is in talks with Google to spend billions of dollars on the Alphabet-owned company’s chips for use in its data centers starting in 2027, according to The Information. This move would solidify Google’s position as a strong competitor to semiconductor giant Nvidia.

The Information, citing sources familiar with the discussions, reported that the talks also include the possibility of Meta leasing chips from Google Cloud starting next year, as part of Google’s broader efforts to encourage customers to adopt tensor processing units (TPUs) for artificial intelligence tasks within their data centers.

This would represent a departure from Google’s current strategy of using only tensor processing units in its data centers and could significantly expand the chip market, putting the company in direct competition for the hundreds of billions of dollars being spent on data center processors to power AI services, according to Reuters.

Some Google Cloud executives have suggested that this strategy could help them capture up to 10% of Nvidia’s annual revenue, a share worth billions of dollars.

The chip deal with Meta, one of Nvidia’s largest customers, which plans to spend up to $72 billion this year, would represent a significant leap forward for Google, which is already one of the biggest beneficiaries of the generative AI boom thanks to increased demand for its cloud services from companies adopting the technology.

Demand for custom chips, such as GPUs, has skyrocketed in recent years as companies seek alternatives to Nvidia’s expensive and limited-supply GPUs. Anthropic announced last month an expanded deal with Google to use up to one million AI chips from the tech giant, worth tens of billions of dollars.

Nvidia’s dominance will require Google to overcome nearly two decades of using Nvidia software that has made it difficult to dislodge its ecosystem. More than 4 million developers worldwide rely on Nvidia’s CUDA software platform to build AI and other applications.