A New Venture, an Old Financing Trick
Crux AI launched in early September 2026 as a joint venture between Blackstone and Alphabet, built to sell AI computing capacity to customers who cannot or will not build their own data centers. Within about a week of its launch, a group of ten banks arranged 22 billion dollars in debt to fund the purchase of the Tensor Processing Units, Google's own AI chip, that the venture needs to operate.
The lead banks include Goldman Sachs, Sumitomo Mitsui Banking Corp, Barclays, BNP Paribas and Bank of Nova Scotia. Alongside the main facility, select lenders are extending a separate 1 billion dollar revolving credit line, and bankers involved expect the short-term debt to eventually be refinanced through investment-grade bonds once Crux AI has a trading history.
Why the Chips Themselves Are the Collateral
What makes this loan unusual is not its size, it is what backs it. The 22 billion dollars is secured against the value of the TPUs Crux AI is buying and against the revenue from its customer contracts, not against Blackstone's or Alphabet's broader balance sheet. That structure is closer to how an airline finances a fleet of planes against the planes themselves than how a hyperscaler has typically funded a data center build.
Blackstone is also putting in 5 billion dollars of equity, and Google is supplying the TPUs plus the software and services layer Crux AI runs on. Neither company has said Crux AI is on its own consolidated balance sheet in the way an internal division would be, which is part of what makes asset-backed debt the more attractive route for scaling fast.
The Deal in Numbers
The figures line up into a clear picture of how a debt-financed AI cloud gets built in 2026.
| Item | Figure |
|---|---|
| Chip-backed loan | 22 billion dollars |
| Revolving credit line | 1 billion dollars |
| Blackstone equity investment | 5 billion dollars |
| Banks arranging the loan | 10, five named publicly |
| Target compute capacity by 2027 | 500 megawatts |
| Time from launch to loan | about 8 days |
What This Means If You Rely on a Smaller AI Vendor
Most AI capacity people actually use today sits on the balance sheets of Microsoft, Amazon, Google or Meta, companies that fund data centers out of enormous operating cash flow and can absorb a slow quarter without a lender calling. Crux AI is a different animal: a standalone venture whose main asset is chips that lose value and usefulness within a handful of years, financed with debt that assumes those chips keep earning throughout their working life.
If your business buys AI inference, training capacity or a managed AI product from a newer or smaller provider, it is worth asking who financed that provider's hardware and how. A company self-funded from an established parent's cash flow can absorb a demand dip quietly. A company carrying billions in chip-backed debt has less room, and a squeeze on it can show up as a sudden price increase, a capacity waitlist or a service the vendor quietly deprioritizes.
The Bottom Line
Debt secured against depreciating silicon is a genuinely new financing layer in the AI buildout, and Crux AI will not be the last venture to use it. The number worth remembering is not 22 billion, it is 8 days: that is how quickly asset-backed lenders are now willing to move once a credible sponsor and an offtake story are in place, and how quickly a customer relationship with one of these ventures can shift if the underlying economics do not hold.
Read next: Europe's Pension Funds May Buy a Circular AI Bet | Five Billion In Defense Money Is Now Chasing GPUs
