Google is reportedly looking to give other cloud providers access to its Tensor Processing Units (TPUs).

The TPUs are Google's in-house AI chips, and have previously been exclusively used by Google itself.

The shift in strategy was reported by the Information. According to the publication, Google has approached several "neoclouds," including Crusoe and CoreWeave, about providing them with access to the TPUs.

London-based Fluidstack is said to have signed a deal already and will deploy the chips in its data center in New York, which is leased from Terawulf. As part of Fluidstack's lease, Google has provided a $3.2 billion backstop and has taken an equity stake in the company of ~14 percent.

DCD has contacted both Fluidstack and Google for comment.

Update -

A Google spokesperson told DCD via email: "

"We have a long history of partnering with data center colocation providers to bring our Cloud technologies closer to customers around the world. Our new partnership with Terawulf is an extension of that strategy as we focus on increasing capacity to meet customer needs.” 

Original story resumes - 

While no deal with CoreWeave to provide TPUs has been noted, in June 2025 it was reported that CoreWeave would provide computing capacity to Google's cloud unit, which would be used under Google's agreement with OpenAI. In July, reports emerged that the Google/OpenAI agreement would include OpenAI using Google's TPUs.

Google's TPUs - or Tensor Processing Units - are chips optimized for AI workloads. The latest generation, dubbed Ironwood, was unveiled in April of this year.

Ironwood also offers 192GB of 7.4Tbps of high-bandwidth memory (HBM) per chip, a 6x increase on Google’s sixth-generation Trillium TPU, in addition to 1.2Tbps bidirectional Inter-Chip Interconnect (ICI) bandwidth, a 1.5x improvement over Trillium.

Other known customers of Google's TPUs include Apple, Anthropic, and Safe Superintelligence.

The likes of Microsoft and Amazon have developed their own AI accelerator chips, but neither has yet made its custom silicon available to third-party cloud providers outside their own data centers.