Sources: a group of 10 banks is giving a $22B loan to Blackstone and Alphabet's cloud venture Crux AI, which aims to provide compute for AI labs, to buy TPUs
A group of 10 banks is providing a $22 billion chip loan to support Blackstone Inc. and Alphabet Inc.'s new cloud venture Crux AI …
Context & Ripple Effects
Blackstone and Google had already formed a US venture to offer customers TPU access, backed by Blackstone's $5 billion initial equity commitment. The reported loan would add bank debt to that TPU-access joint venture, rather than leave its expansion dependent on sponsor equity alone.
The structure follows a widening TPU-finance pattern: Google had assembled a reported financing program for Anthropic, including TPU-linked commitments, and Apollo and Blackstone had finalized a $35 billion TPU leasing package for Anthropic. Crux would make Blackstone a repeat financier and owner-side partner in that compute build-out.
First-order effects
- If finalized, the reported $22 billion facility gives Crux AI dedicated borrowing capacity to acquire TPUs, scaling the venture's ability to supply compute to AI labs.
- The 10 lending banks would take direct exposure to a TPU-backed cloud venture, while Blackstone and Alphabet would combine their equity partnership with a larger debt layer.
Second-order effects
- Crux's debt-funded TPU purchases reinforce the financing route Google has used around Anthropic, making access to large bank and private-capital pools a more important input to serving AI-lab demand than sponsor equity alone.
- Blackstone's participation across Crux and the Anthropic TPU lease financing concentrates its role as both capital provider and infrastructure partner, raising the competitive importance of specialized compute-finance structures for other cloud ventures.
Third-order effects
- If repeated across providers, TPU deployment shifts toward a capital-stack model in which banks, private credit, strategic chip suppliers and infrastructure sponsors share the cost and risk of supplying AI compute.
- The reported loan is another data point in the financialization of AI infrastructure: scaling compute increasingly depends on whether equipment-backed projects can be structured for outside lenders, not solely on cloud operators' balance sheets.
The trend: AI compute is being turned into a financeable infrastructure asset, with cloud ventures pairing strategic technology partners and alternative-asset sponsors with increasingly large debt facilities.