CoreWeave will reportedly provide some of the computing capacity to OpenAI under the recent deal between the AI company and Google.

As reported by Reuters and citing sources "familiar with the matter," CoreWeave will sell its computing capacity to Google's cloud unit, which will then sell the capacity on to OpenAI.

The news follows reports that Google Cloud had signed a capacity agreement with OpenAI, despite Google's DeepMind AI unit being in direct competition, and ChatGPT posing a threat to Google's search business.

In addition to the Matryoshka doll leasing structure with CoreWeave, Google will also directly provide OpenAI with compute capacity via its own cloud platform.

CoreWeave, OpenAI, and Google declined to comment to Reuters.

CoreWeave previously signed a deal with Google in Q1 of 2025, with previous reports noting Google was in talks to rent Nvidia Blackwell chips from the neocloud.

The leasing-to-then-release structure has previously been seen with OpenAI, Oracle, and Microsoft.

OpenAI was leasing servers from Microsoft, which was then leasing them from Oracle in June 2024. At the time, however, OpenAI held an agreement with Microsoft to solely use the cloud provider's computing resources to train its LLMs. This agreement enabled OpenAI to technically continue naming Microsoft as its exclusive cloud partner.

OpenAI has since diversified its cloud partners.

With the AI company having no such agreement with Google Cloud, the decision to release compute from CoreWeave is presumably driven by an insatiable need for capacity. According to Reuters' sources, OpenAI had previously sought a deal with Google Cloud but was blocked by its lock-in with Microsoft.

Other partnerships OpenAI is using to expand its compute capacity include the $500bn Stargate project, announced at the start of this year in partnership with SoftBank, Oracle, and Abu Dhabi's MGX. The first campus will be in Abilene, Texas, for which Oracle is supposedly set to purchase around 400,000 Nvidia GB200s.

Beyond the US, OpenAI is exploring developing Stargate campuses in other locations, including in the UAE, and OpenAI’s chief strategy officer is currently touring Asia Pacific, including India, Japan, South Korea, Australia, and Singapore - with the trip in part to hear potential data center investment incentives. In Europe, OpenAI is reportedly in discussions with the UK, France, and Germany.

OpenAI also signed a $4bn cloud deal with CoreWeave in May 2025, building on an earlier $11.9bn deal with the neocloud.

This week, financial services firm TD Cowen said that it believed Oracle was seeking around 5GW of data center capacity in the US for OpenAI, separate from the Stargate project. This was seemingly confirmed, though not the specific number, in Oracle's recent earnings call, with CEO Safra Catz stating that OpenAI was contributing to its significant cloud growth despite Stargate still being "in formation."