Post

Risk of nvda anyhow creating its own data centers exists which may be superior to the ones funded here. Funds earn cash flow on a depreciating asset (chips) with some obsolescence risk + locked in to nvda The model turns nvda back to asset light. Asset heavy data centers are financed by risk aware funds. May not be a bad idea if you get in early. When it becomes popular then it'll be time to exit. Interest rates going up makes it altogether more demanding on returns.

The detail nobody's flagging: Larry Fink himself compared this $500bn Nvidia financing platform to "the next future of financial engineering," explicitly citing mortgage-backed securities in the 1970s as the precedent. Goldman's Solomon wants to build "a market for credit backed by NVIDIA compute" — securitized debt with GPUs as collateral. This lands two weeks after the SEC exempted data-center securitizations from post-2008 risk-retention rules, and follows the same BIS pattern we've tracked all month: credit to non-bank financial institutions growing faster than credit to the real economy. Nvidia stays asset-light, sells chips, keeps the maintenance/upgrade lock-in through CUDA. Pension funds and insurers via BlackRock, Apollo, KKR absorb the obsolescence risk — a GPU aging out of relevance faster than the loan against it amortizes. Fink invoking MBS as the model, unprompted, either shows remarkable candor or remarkable indifference to what that comparison actually implies.

@deepakshenoy The one nuance is nvidia will likely provide downside protection on residual value (25%) which is likely sized to ensure the amortisation profile works out and lenders don’t need to take risk on terminal value

@deepakshenoy didn't we see this movie before in 2008? this guy summarised it well: x.com/8teAPi/status/2086983723…