Google is teaming up with private equity giant Blackstone on a new U.S.-based joint venture that will sell access to Google’s Tensor Processing Units (TPUs) as a compute-as-a-service product, backed by an initial $5 billion equity commitment from Blackstone. The partners say they are targeting roughly 500 megawatts of AI compute capacity online by 2027 and have tapped long‑time Google infrastructure leader Benjamin Treynor Sloss to run the operation. The deal opens a new commercial path for Google’s custom chips beyond Google Cloud and pushes Blackstone deeper into the high‑power compute arena.
According to Blackstone, the firm is putting up the initial $5 billion and expects the first 500 MW of capacity to be available in 2027. “We see a generational opportunity to invest capital at scale building AI infrastructure,” Jon Gray, Blackstone’s president and COO, said in the company’s announcement. The release also confirms Sloss as CEO and pitches the venture as marrying Blackstone’s infrastructure capital with Google’s TPU technology.
What Google will provide
Google says it will deliver the TPUs along with the software and services stack, so customers can tap Google-designed AI silicon through a separate commercial channel from Google Cloud. In Google's write-up, Google Cloud CEO Thomas Kurian said the partnership helps meet growing demand for TPUs, which are optimized specifically for efficiency and performance in the AI era. Google frames the structure as a way to speed up customer access to accelerator hardware without putting additional capacity on Google Cloud’s balance sheet.
Scale and financing
The timing lands as demand for accelerated compute continues to outrun current supply, with Reuters reporting on the creation of a U.S.-based AI cloud company underpinned by Blackstone’s equity. Other outlets, citing Bloomberg, have said the eventual buildout could reach roughly $25 billion once debt financing is included, putting it in the same ballpark as the greatest hyperscale efforts. Industry watchers say the setup lets Google widen the distribution of its TPUs while leaving the heavy lifting on capital and construction to an infrastructure specialist.
Local and market implications
For Bay Area readers, the partnership is another reminder that Silicon Valley still sits at the center of strategic compute deals, even as much of the physical infrastructure is expected to chase cheaper power in other regions. The news was flagged locally by the Silicon Valley Business Journal, which pointed out that the agreement lands amid a wave of TPU supply arrangements. Public-market traders did not wait around either: specialist AI-compute players such as CoreWeave and similar providers saw their shares react to the announcement, according to market coverage. Investing.com reported early trading that showed pressure on some niche competitors.
What to watch next
Several questions now hang over the new venture: which enterprises and AI labs will sign on for capacity, how TPU pricing will stack up against GPU-based options, and where the first campuses will actually land as the partners aim for their 2027 target. For any Bay Area buildout, local permitting, transmission headroom and utility deals will be key, while the rollout pace will hinge on Blackstone’s ability to phase capital and nail down power. If the effort scales to the levels cited in business reporting, it could significantly influence both supply and pricing for top-tier AI compute.
For startups and larger customers alike, the move promises more ways to get TPUs, along with fiercer competition for both capacity and the engineering talent needed to put it to work. Expect scrutiny of early customer wins, regulatory filings and site decisions in the weeks ahead.
