Nvidia Partners With Six Financial Institutions on 500 Billion Dollar AI Financing
Nvidia has signed memorandums of understanding with six of the world’s largest financial institutions to create financing platforms intended to mobilise more than five hundred billion dollars of third party capital for the construction of artificial intelligence infrastructure, the company announced on Monday.
The partners are Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR. Under the agreements Nvidia will work with each to assemble dedicated pools of capital, described by the company as being available at significant scale and at attractive rates to its own customers, who use the money to buy processors, build data centres and secure long term electricity supply. The company said the platforms would be independent, and that the partnerships remain subject to execution of the final agreements.
The distinction between mobilising capital and committing it is the substance of the announcement. Nvidia is not investing five hundred billion dollars. It is convening six firms that are expected to raise that sum from their own investors over time. The company’s stated purpose is to turn its compute and full stack infrastructure into what it calls an investable asset class, with long duration revenue linked to usage.
| Partner | Stated role or existing tie to Nvidia |
| Apollo | Long term capital base; frames the buildout as part of a broader industrial renaissance |
| BlackRock | Deepens an existing relationship through the AI Infrastructure Partnership |
| Blackstone | Describes itself as an existing large investor across the Nvidia ecosystem |
| Brookfield | Existing strategic partner; compute named a core pillar of its AI infrastructure strategy |
| Goldman Sachs | Investment and distribution roles; aims to create a market for credit backed by Nvidia compute |
| KKR | Nvidia is a founding investor in KKR’s Helix Digital Infrastructure |
Jensen Huang, founder and chief executive of Nvidia, said the company had reached an important milestone, having begun by building chips and now helping to create a new class of productive, investable infrastructure. He argued that in artificial intelligence compute is revenue, and that Nvidia’s compute is suited to the role because it is broadly adopted, flexible across models and workloads, transferable between customers and operators, and continuously improved through the company’s software, which extends its useful life. He said the platforms would help customers reach scarce compute at scale.
The financiers framed it as an asset class question. Jim Zelter, president of Apollo, called modern compute a scarce, mission critical asset class with compelling investment characteristics. Larry Fink, chairman and chief executive of BlackRock, said the buildout would require unprecedented investment and a skilled workforce. Jon Gray, president and chief operating officer of Blackstone, pointed to the demand Nvidia has created for its own compute. Bruce Flatt, chief executive of Brookfield, said compute was becoming the essential layer of infrastructure. Joe Bae and Scott Nuttall, co chief executives of KKR, said that in digital infrastructure delivery rather than ambition is the hard part.
The most specific description of a structure came from Goldman Sachs. David Solomon, its chairman and chief executive, said the firm was looking to create a market for credit backed by Nvidia compute. The company’s announcement does not say that every platform will be built the same way, and no terms, pricing or collateral arrangements were disclosed for any of them.
The arrangement reflects a shift in who is paying for the technology cycle. The capital required to build and power data centres has outgrown what the operators can fund from cash flow or public markets, and alternative asset managers have been the willing lenders, tapping institutional and insurance capital, according to CNBC. Several of these firms have already arranged debt and equity for companies building large scale artificial intelligence capacity.
Why it matters: A hardware manufacturer convening the financing that its customers will use to buy its hardware is an unusual arrangement, and the interesting question is what it says about demand rather than what it says about supply. If buyers could fund purchases on their own balance sheets, the platforms would not be needed. The company’s answer is that the constraint is the cost of capital rather than the appetite, and Goldman Sachs has put the sharpest version of the proposition on the table by talking about a market for credit backed by Nvidia compute. That raises a question nobody has yet answered publicly, which is what a processor is worth as security and for how long. Nvidia’s argument is that its software keeps the hardware useful for longer than depreciation schedules assume, and that argument has not been priced. Nothing is committed yet either. These are memorandums subject to final agreements, and the sum is a target for capital to be raised over time, not money in place.
Looking ahead: The first signal is the conversion of the memorandums into executed agreements, and then the first fund closes and their terms, because pricing will reveal what lenders actually think a processor is worth as security. Independence is the second test, given that the platforms are meant to be underwritten independently while being convened by the manufacturer. Nvidia reports quarterly results later this month, when any accounting treatment or commitment attached to the platforms would ordinarily be disclosed.
Sources: NVIDIA Corporation press release, 10 August 2026; CNBC.
