Nvidia signed memorandums of understanding on 10 August with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR to mobilise more than $500 billion of outside capital for AI data centres. Each firm operates its own independent financing platform, lending to Nvidia’s customers so they can build facilities and buy Nvidia hardware without funding it from their own balance sheets. Jensen Huang told CNBC his chips are an investable asset.
The mechanism is what is new. These platforms treat GPU compute as collateral, the way lenders already treat power plants, toll roads and commercial property. That reframes a depreciating piece of consumer-cycle silicon as an infrastructure asset with a predictable yield, and it is the step that unlocks pension and insurance money which cannot buy chips but can buy infrastructure debt.
The problem it solves is real. Analysts have identified funding rather than fabrication as the binding constraint on the AI buildout, because the hyperscalers and labs that want capacity have been reaching the limits of what their own balance sheets and existing lenders will carry. Moving the financing off Nvidia’s books and onto six institutional platforms addresses that directly.
Two things sit underneath it. Collateral only works if there is a resale or re-lease market at a known price, and GPU residual values have never been tested through a demand downturn, with each new generation cutting what the last one is worth. And Nvidia is helping arrange the financing for purchases of Nvidia products, which supports demand for its own hardware in a way that makes the demand harder to read from outside. These are memorandums, not executed agreements, and the deals are expected to reach market over the coming months.
