- Lead. Nvidia announced on August 10 that it has signed memorandums of understanding with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR to establish AI compute infrastructure financing platforms targeting over $500 billion in third-party capital.
- Fact. Nvidia CEO Jensen Huang, who described his chips as “fungible and transferable,” is positioning GPU compute as a new financeable asset class — comparable to commercial real estate or toll roads — backed by long-term AI demand.
- Stake. Blackstone and KKR each rose more than 6% on August 11; Apollo surged more than 6%. The market read the announcement as a signal that private capital is about to flow into AI hardware at a scale that makes the six firms significant beneficiaries.
The announcement, published on Nvidia’s newsroom on August 10, represents an attempt to solve one of the central bottlenecks in AI buildout: the gap between how much compute hyperscalers and frontier labs need and how much capital they can access through conventional equity and debt. Nvidia’s proposed solution is to treat its hardware as the collateral layer, creating dedicated lending platforms that extend credit against defined pools of Nvidia compute at what the company describes as “attractive rates.”
CEO Jensen Huang told media he personally approached all six firms and “none turned him down” — a formulation intended to signal institutional credibility as much as business momentum. He described Nvidia compute as “broadly adopted, flexible across models and workloads, fungible and transferable,” the qualities that make it suitable as financial collateral rather than bespoke hardware tied to a single customer or architecture.
The Capital Stack
Nvidia is backstopping 25% of the financing across the platforms, with the bulk of credit exposure carried by the six alternative asset managers and banks. The structure echoes how private credit markets have approached infrastructure classes — data centres, fibre networks, cell towers — by securitising income-generating assets that trade on long-term demand rather than quarterly earnings cycles. Jim Zelter of Apollo described modern compute as “a scarce, mission-critical asset class with compelling investment characteristics,” a framing that positions AI hardware alongside pipelines and port terminals in private capital’s infrastructure allocation.
Morgan Stanley estimates major hyperscalers will spend approximately $3.5 trillion on AI infrastructure between 2026 and 2028; Apollo has projected total AI infrastructure investment could eventually exceed $8 trillion. The capex commitments already disclosed by Alphabet, Amazon, and Microsoft confirm that demand is real — the constraint has been sourcing capital at rates that make the economics work for buyers who cannot fund GPU purchases from operating cash flow alone.
Partner Statements
Each of the six firms issued statements alongside the Nvidia announcement. Larry Fink of BlackRock described the deal as deepening the firm’s relationship with Nvidia “including through the AI Infrastructure Partnership.” Bruce Flatt of Brookfield called compute “the essential layer of infrastructure and a core pillar of the Brookfield AI infrastructure strategy.” David Solomon of Goldman Sachs said Nvidia’s “full-stack platform is in high demand and uniquely positioned at the center of that global buildout.” KKR co-CEOs Joe Bae and Scott Nuttall offered the most cautious framing: “Delivery, not ambition, is the hard part.”
What the MOUs Mean
The agreements signed are memorandums of understanding, not final contracts. Each platform’s parameters — interest rates, eligible collateral, loan-to-value ratios, repayment terms — will be set through individual negotiations, with final terms subject to execution. The announcements confirm intent and alignment, not deployed capital. The harder test is whether AI compute, once used as loan collateral at scale, retains the residual value assumptions that make the lending economics work — or whether rapid model iteration devalues hardware faster than lenders can recover principal.