Banks line up $22 billion loan for Crux AI, the Blackstone-Google TPU cloud venture

SentiSense · Published · Updated

A group of 10 banks is providing a $22 billion loan to Crux AI, the cloud venture Blackstone and Alphabet's Google announced in May, Bloomberg reported on September 16 and Reuters confirmed the same day. The debt will be used to buy Google's Tensor Processing Units and is secured against the chips themselves and against Crux AI's customer contracts. Goldman Sachs, Sumitomo Mitsui, Barclays, BNP Paribas and Bank of Nova Scotia are among the lenders. The venture buys and operates Google-made chips as cloud infrastructure rather than designing hardware of its own.

A group of 10 banks has lined up a $22 billion loan for Crux AI, the cloud venture that BX and GOOGL subsidiary Google announced in May, Bloomberg reported on September 16 and Reuters carried the same day. Goldman Sachs, Sumitomo Mitsui Banking Corp, Barclays, BNP Paribas and Bank of Nova Scotia are among the lenders named in the reporting.

The structure is the interesting part. Reuters reports that the debt will be used to purchase Google TPUs and that it will be backed by the value of those chips and by Crux AI's customer contracts. That is asset-backed lending against silicon and contracted revenue rather than a corporate facility drawn on either sponsor's balance sheet, and it puts a price on AI accelerators as collateral at a moment when the market has few reference points for what that hardware is worth over a loan's life.

Crux AI itself is a buyer of chips, not a builder of them. Reuters describes it as a cloud business serving AI labs, technology companies, enterprises and governments, with Google supplying its custom Tensor Processing Units along with software and services. Blackstone's contribution is capital and data-center buildout: at the venture's launch in May the firm committed $5 billion of equity and set a target of 500 MW of capacity online in 2027.

What to watch is whether this financing shape spreads. If lenders are willing to underwrite $22 billion against TPUs and offtake contracts, the capital available for AI compute buildouts may be considerably deeper than balance-sheet capacity alone would suggest, which could matter for every hyperscaler and neocloud competing for the same power and hardware. The risk runs the other way too: collateral whose resale value depends on a fast-moving chip roadmap is harder to mark than the real estate and turbines that usually sit behind infrastructure debt. Neither sponsor has disclosed the full lender syndicate or the loan's terms.

Powered by SentiSense - Intelligent Market Analysis