NVIDIA announced on August 10, 2026 that it had signed memorandums of understanding with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR to establish independent compute financing platforms intended to mobilize over 500 billion dollars of third-party capital for the buildout of AI infrastructure over time. Under the partnerships NVIDIA will work with the six firms to create dedicated pools of capital at significant scale and at attractive rates for NVIDIA customers, aimed at frontier AI labs, enterprises and AI clouds across its ecosystem. The announcement states that the partnerships remain subject to execution of final agreements.
The structuring argument NVIDIA makes is that its compute qualifies as a financeable asset in its own right. The company describes NVIDIA compute as providing the lowest token cost, highest revenue and longest life, and points to fungibility as the credit feature: hardware that is flexible across models and workloads and transferable between customers and operators, with useful life extended by continuing CUDA software improvements. Jensen Huang put it as a shift in what the company sells, saying NVIDIA began by building chips and is now helping create a new class of productive, investable infrastructure in AI factories, and that in AI, compute is revenue.
The lenders framed it as an asset class question rather than a vendor relationship. Apollo president Jim Zelter called modern compute a scarce, mission-critical asset class. Goldman Sachs chairman and chief executive David Solomon described the opportunity as creating a market for credit backed by NVIDIA compute. KKR co-chief executives Joe Bae and Scott Nuttall noted NVIDIA is a founding investor in Helix Digital Infrastructure, and BlackRock chairman and chief executive Larry Fink tied the platforms to the existing AI Infrastructure Partnership.
The consequence to watch is who carries the risk. Until now the AI buildout has been funded mainly off hyperscaler balance sheets, neocloud debt raises and vendor equity stakes, all of which concentrate exposure in a small number of operators. Third-party platforms that lend against compute itself move that exposure to long-duration institutional capital and let operators expand without pledging their own balance sheets, which lowers the cost of adding capacity. The same mechanism also creates a new channel through which a downturn in token demand would transmit into credit markets rather than staying inside technology equity, and the phrase “credit backed by NVIDIA compute” makes the residual value of accelerators a question with financial consequences beyond the buyers.