For most of the AI boom, Nvidia had the easiest job in technology: build the best chips, then wait for the orders to arrive. This month it took on a harder one. Alongside six of the biggest names in finance, including Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR, the company unveiled a platform designed to mobilise as much as $500 billion in outside capital to fund the data centres that AI runs on. Each partner will vet projects on its own, and Nvidia has signalled it may provide residual backing of up to 25 percent on individual investments. The chipmaker, in other words, is no longer just selling the shovels. It is helping arrange the loans that pay for them.
Jensen Huang framed the move as a milestone rather than a rescue. “This is really the first time that technology chips have become an investable asset class,” he told CNBC. “These are revenue-generating assets now. They’re productive, they’re long-lived, they’re fungible, they’re flexible.” The pitch is that a GPU cluster is less like a gadget that ages into obsolescence and more like a toll road or a power plant, an asset worth financing over years. Morgan Stanley, which kept its Overweight rating, liked the structure precisely because the money comes from third parties: it eases the worry that Nvidia is quietly funding its own customers, while a revenue-sharing arrangement could lift the company’s 2029 earnings by more than 10 percent.
Not everyone reads it so kindly. Michael Burry, the investor best known for betting against the 2008 housing market, has disclosed short positions against Oracle and the cloud provider Nebius, and singled out Nvidia’s financing push and its “complex web of private credit and asset-backed securities.” His worry is circularity: if buyers increasingly lean on debt to purchase Nvidia hardware, and Nvidia is helping supply or backstop that debt, then demand starts to look self-referential. Sales that are ultimately financed by the seller can flatter a boom right up until the financing dries up.
There is a quieter data point that suggests Nvidia sees the risk too. A Wall Street Journal report, widely discussed this week, says the company pared back its financial guarantee on a major OpenAI data centre project, from an initial figure near $250 billion to roughly $150 to $160 billion. Observers described it as a calculated move to limit exposure should the resale value of GPUs fall. So the same company is leaning in and hedging at once: building a marketplace to unlock half a trillion dollars of other people’s money, while trimming the amount of its own money on the line.
All of this lands nine days before earnings. Nvidia reports its fiscal second quarter on 26 August, and the stock has already run up roughly 19 percent since late July, which sets a punishing bar. Last quarter it booked $81.6 billion in revenue, up 85 percent, with data centre sales alone at $75.2 billion. The financing platform is a bet that the buildout has years left to run and simply needs more capital plumbing to get there. The bear case is not that AI demand is fake. It is that when the biggest seller becomes one of the biggest financiers, the line between real demand and manufactured demand gets harder to see, and that is exactly the line the next quarter will be judged against.