NVIDIA lines up six asset managers behind $500B to make GPUs a financeable asset

NVIDIA said on August 10 that it has signed memorandums of understanding with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR to establish compute financing platforms intended to mobilize more than $500 billion of third-party capital for AI infrastructure over time ^[1][2]. What is being attempted is not a sale but a reclassification: NVIDIA wants its hardware treated as an investable asset class producing long-duration, usage-linked revenue, which shifts the binding constraint from who can afford GPUs to who will underwrite the cash flows they generate ^[4][18].

The pitch is explicit. NVIDIA describes its compute as offering the lowest token cost, the highest revenue and the longest life, supported by a set of offtakers built on CUDA ^[5]. Jensen Huang's formulation is that "in AI, compute is revenue," and that the asset is fungible and transferable across customers and operators, with CUDA software extending its useful life and improving its economics over time ^[6]. That is a residual-value argument written for credit committees rather than developers. If a fleet can be re-let to another operator when a tenant defaults, loss-given-default assumptions improve and debt tenor can stretch beyond the tenant's contract.

Three things are absent. NVIDIA did not disclose financial terms, commitments by individual firms, or a timetable for deploying the $500 billion ^[10]. Memorandums of understanding are not capital ^[2]. And the underwriting is less arm's length than the wording implies: the release frames the goal as having long-term capital providers "independently underwrite AI infrastructure" ^[7], while Huang said on X that NVIDIA has the option to backstop up to $125 billion, or 25% of the potential deals ^[8]. Even at that ceiling, up to $375 billion would sit on offtaker credit and residual values with no vendor guarantee ^[9]. A supplier offering to stand behind a quarter of the financing of its own product is vendor financing with a cap, and it is the part of the structure that most affects how much risk actually leaves Santa Clara.

For scale, Reuters reports combined Big Tech AI outlays are set to surpass $730 billion this year ^[11], so a $500 billion pool spread over an unspecified period is roughly two-thirds of one year of that spending ^[12]. Its function is not to replace hyperscaler capex but to fund the buyers who are not hyperscalers: frontier AI developers, enterprises, governments and cloud providers without investment-grade balance sheets ^[18].

The financiers are not neutral observers. Apollo president Jim Zelter called modern compute "a scarce, mission-critical asset class with compelling investment characteristics" ^[14]. Larry Fink said the arrangement deepens BlackRock's existing relationship with NVIDIA, including through the AI Infrastructure Partnership ^[15]. Blackstone's Jon Gray said the firm continues to be an enormous investor globally across the NVIDIA ecosystem ^[16], and Brookfield's Bruce Flatt described compute as a core pillar of the firm's AI infrastructure strategy ^[17]. Prior exposure explains the confidence and also means the same balance sheets appear on both sides of the ecosystem. The Financial Times reported the plan first, with Reuters confirming it ^[13].

Watch for the memorandums converting into definitive agreements with disclosed per-firm commitments and a deployment schedule ^[10]. Watch whether the platforms lend against contracted offtake or against assumed residual value, and what depreciation schedule and lease tenor they adopt relative to contract length. And watch for any disclosure on the $125 billion backstop option, which determines whether this is genuinely third-party risk ^[8].