Source: CNBC — 2026-08-10
Summary
Nvidia signed MOUs with six major Wall Street firms — Apollo Global Management, Blackstone, BlackRock, Brookfield, Goldman Sachs, and KKR — to mobilize over $500 billion in third-party capital, letting hyperscalers, frontier AI labs, and enterprises finance data-center buildouts and GPU purchases without straining their own balance sheets. The structure treats AI chips and data centers as a new collateralizable asset class, comparable to how commercial real estate or toll roads get financed, with BlackRock's Larry Fink comparing it to the birth of mortgage-backed securities in the 1970s. Nvidia CEO Jensen Huang said he approached only these six firms and none declined.
Key Takeaways
- The core mechanism: AI infrastructure (chips, data centers) becomes collateral for structured financing, moving AI capex off tech-company balance sheets and into credit markets — the same shift real estate and infrastructure went through decades ago.
- Six firms are involved — Apollo Global Management, Blackstone, BlackRock, Brookfield, Goldman Sachs, and KKR — spanning private equity, asset management, and investment banking, not a single financing partner.
- BlackRock's Larry Fink's mortgage-backed-securities comparison is a direct signal that Wall Street sees GPU/data-center debt as headed toward the same securitization playbook that reshaped housing finance in the 1970s.
- Jensen Huang's claim that he approached exactly six firms and none declined suggests strong appetite from institutional capital for AI infrastructure exposure, not a hard sell.
- This decouples AI buildout speed from any single company's cash flow or balance sheet capacity — capital-constrained labs and hyperscalers can now scale compute by tapping structured credit markets instead.
Reel Script
Hook (17s)
Nvidia just got six of the biggest names in finance to agree to move half a trillion dollars into AI infrastructure — not by buying chips themselves, but by turning GPUs into something you can borrow against, like a mortgage.
Core Concept (85s)
Building AI data centers is absurdly expensive — GPUs, power, cooling, real estate — and right now most of that gets paid for out of tech companies' own cash and balance sheets. That's a bottleneck: even a company as large as a hyperscaler can only stretch its balance sheet so far before lenders and shareholders get nervous. Nvidia's answer is to make AI infrastructure "financeable" the way real estate already is. When you buy a house, the bank doesn't just trust your paycheck — it uses the house itself as collateral, and then often bundles thousands of these mortgages together and sells them to investors as securities. Nvidia signed agreements with six major financial firms — Apollo, Blackstone, BlackRock, Brookfield, Goldman Sachs, and KKR — to build the equivalent system for GPUs and data centers: treat the chips and buildings as collateral, structure loans and credit around them, and pull in outside investor capital instead of relying purely on tech company cash. That's the mechanism — turning a depreciating tech purchase into a financeable, collateralizable asset class.
Hands-On (100s)
The headline number is $500 billion-plus in third-party capital these six firms are set to help mobilize — money that doesn't come from Nvidia, Microsoft, or OpenAI's own balance sheets, but from investors putting capital into structured financing vehicles backed by chips and data centers. The comparison that matters here came from BlackRock CEO Larry Fink, who likened this moment to the birth of mortgage-backed securities in the 1970s — when banks first figured out how to bundle individual home loans into tradeable securities, unlocking a massive new pool of capital for housing. Nvidia is betting the same trick works for compute: bundle GPU and data-center financing into something institutional investors can hold, trade, and price like they already do with commercial real estate or toll-road debt. And notably, Jensen Huang said he went to exactly six firms for this — and all six said yes, which tells you the appetite for this kind of AI-infrastructure exposure is already there on Wall Street's side.
Takeaway (24s)
This is the moment AI capex stopped being a tech-industry balance-sheet story and became a credit-markets story — expect data-center debt to start trading the way real estate debt does. If you're in fintech or infra, watch how these structured vehicles get priced; that's where the next wave of AI-adjacent deal flow shows up.