Hmmmm.
This is the problem that SF Compute solves, by building a physically settled forwards market for compute. We make it possible for lenders to use the GPU hour as collateral, on a liquid book, rather than the GPU cluster, which is non-fungible and requires a servicer that might blow up right at the moment they need it the most. If a lender gets a GPU hour as collateral, the credit risk of the customer matters less; if the two person AI lab in a San Francisco Victorian fails to make their payments, the lender has a real path to selling the collateral. That makes startups possible at all, a path that seems increasingly difficult in AI.