Nvidia recruits Wall Street to finance AI infrastructure

Nvidia is partnering with six Wall Street firms in a financing push it says could mobilise more than $500 billion for artificial-intelligence infrastructure, according to Reuters. The company says the arrangements will create platforms that let third-party investors fund data-centre buildouts and treat AI compute as an asset class. Nvidia’s chief executive Jensen Huang wrote on X that the company could backstop up to $125 billion of potential deals.

The roster—Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR—reads like a list of institutions that specialise in turning long-lived, cash-generating projects into investable products. Nvidia’s pitch is that “AI factories” can be financed the way pipelines, telecom towers or aircraft fleets are: with dedicated pools of capital and predictable contracts. Reuters notes that Nvidia did not disclose financial terms, commitments by individual firms, or a timetable for deploying the headline figure, leaving the announcement as much about creating a market category as about any single cheque.

The timing reflects how quickly AI spending has shifted from software budgets to power, land, chips and cooling. Reuters cites big tech signalling that AI outlays are expected to surpass $730 billion this year, with Nvidia counting Google, Amazon, Microsoft and Facebook owner Meta among its customers. When demand is constrained by scarce compute, the bottleneck is no longer model talent but the ability to secure capacity—often years in advance—on terms that don’t blow up a company’s balance sheet. A financing wrapper lowers the upfront pain for customers while keeping the hardware pipeline full for the supplier.

For the financiers, “compute” becomes something closer to a toll road: a large initial build, followed by contracted usage and refinancing opportunities. KKR co-chief executives Joe Bae and Scott Nuttall told Reuters that compute has become critical infrastructure, language that tends to travel well in boardrooms and with regulators. Once compute is framed as infrastructure, it becomes easier to argue for preferential treatment in permitting, power allocation and industrial policy—and harder for outsiders to compete without the same capital-market plumbing.

The structure also concentrates leverage in a small number of gatekeepers. If AI capacity is bundled into funds and platforms run by the same handful of asset managers, pricing and access will be shaped by contract terms that are largely private, even as the resulting systems increasingly mediate public life. Nvidia, valued at $5.3 trillion according to Reuters, is positioning itself not just as a chip seller but as the standard-setter for how compute is measured, financed and rationed.

Nvidia says the goal is to help customers access scarce compute at scale. The company has not said when the first financed “AI factories” will break ground or who will get the first call on the capacity.