Google is taking steps to increase the addressable market for its proprietary artificial-intelligence accelerators by financially supporting data-center partners that can deliver compute to a wider set of customers.
People familiar with Googles plans told the Wall Street Journal that the company is exploring increased funding for partners that can broaden access to its tensor processing units, or TPUs, which are its in-house AI chips. One of the potential moves under consideration is an investment of roughly $100 million in cloud-computing company Fluidstack as part of a transaction that values the startup at about $7.5 billion, according to those people.
Fluidstack is identified by those sources as one of a growing cohort of so-called neocloud firms that provide computing services to AI customers. The reported investment and other possible financial commitments are intended to support Fluidstacks growth and to encourage additional computing providers to adopt Googles TPUs, the people said.
In addition to the Fluidstack discussions, Google has reportedly held talks about expanding financial commitments to other data-center partners that could increase demand for its accelerators. Those conversations are described by people familiar with the matter as part of a broader push to make TPU-based compute more widely available beyond Googles own cloud and internal users.
Within Google, managers at the cloud division recently revisited a long-running internal debate over whether to reorganize the TPU team into a stand-alone unit. Proponents say that creating a separate organizational entity could open up additional investment pathways, including the possibility of bringing in outside capital. The people familiar with the discussions framed such a move as a way to expand investment opportunities tied to TPUs.
Google, however, pushed back on speculation about structural change. A company representative stated there are no plans to restructure the TPU unit, according to the person familiar with the messaging.
The discussions described by the people contacted for the report are presented as exploratory and focused on expanding TPU adoption through external partnerships and selective capital deployment. Observers inside and outside the company would likely view the initiatives as measures to increase the competitive reach of Googles AI hardware, though the people cited did not provide a timetable or commitments beyond the conversations and the Fluidstack discussions.
Summary
Google is exploring ways to broaden the customer base for its tensor processing units by providing financial backing to data-center operators. The company is reportedly in talks to invest approximately $100 million in Fluidstack, valuing the startup at about $7.5 billion, and has discussed expanding commitments to other partners. Internally, Google revisited a debate about reorganizing the TPU team into a stand-alone unit, though a Google spokesperson said there are no plans to do so.