Analysis
Nvidia signed memorandums of understanding with Apollo Global Management, Blackstone, BlackRock, Brookfield Asset Management, Goldman Sachs and KKR to build financing platforms that could mobilize more than $500 billion in third-party capital for hyperscalers, frontier AI labs and enterprises buying Nvidia hardware and building data centers, according to CNBC. CEO Jensen Huang told CNBC in a separate interview that "this is really the first time that technology chips have become an investable asset class," arguing that Nvidia's GPUs are now "revenue-generating assets" that are "productive, long-lived, fungible and flexible" enough to be financed the way toll roads or power plants are.
Why asset managers are underwriting GPUs
The pitch is straightforward: rather than hyperscalers funding every data center buildout off their own balance sheets, institutional credit, insurance capital and private equity can underwrite the GPUs and the buildings around them, using the hardware and its contracted revenue as collateral. That's the same basic structure already visible elsewhere in AI infrastructure financing:
“- Oracle -- roughly $72 billion in Stargate partner debt across data centers in Michigan, Texas, Wisconsin and New Mexico.”
- CoreWeave -- $8.5 billion investment-grade GPU-backed financing facility, closed March 2026, anchored by Meta's contracted backlog.
- Oracle -- roughly $72 billion in Stargate partner debt across data centers in Michigan, Texas, Wisconsin and New Mexico.
- Nvidia / OpenAI -- Nvidia has already committed up to $250 billion to backstop OpenAI's Ohio data center buildout.
- This platform -- the new MOUs target more than $500 billion, meant to standardize and scale that same vendor-financed structure across more customers.
Who's actually on the hook
Wall Street's willingness to underwrite this at $500 billion scale is a bet that GPU demand stays durable enough that lenders get repaid even if any single AI lab's business model wobbles. Apollo, Blackstone, BlackRock, Brookfield, Goldman and KKR collectively manage trillions in assets, and pension funds, insurers and sovereign wealth vehicles are the ultimate capital behind private credit funds -- meaning retirement money is now indirectly exposed to whether AI data center buildouts generate the returns Huang is promising.
The circularity risk
The risk is the one every skeptic of AI capex has been flagging for over a year: circularity. Nvidia sells chips to hyperscalers and AI labs, some of whom are financed in part by capital Nvidia helped arrange or backstop, which then buys more Nvidia chips. If AI infrastructure utilization or pricing power disappoints, the losses don't stay contained to equity investors -- they flow into credit markets that are supposed to be more senior and protected. Huang's framing of chips as "infrastructure, like electricity, like the internet" is a useful analogy for financiers, but electricity and internet infrastructure have decades of demand history behind them; GPU clusters optimized for today's model architectures have none.
What it means for LPs and GPs
For founders and GPs, the counterweight is worth sitting with: every one of these financing structures assumes GPU depreciation curves and utilization rates hold roughly where they are today. Nvidia's own hardware cadence -- new chip generations arriving faster than the multi-year loan terms backing older clusters -- is itself a risk to the collateral value these lenders are underwriting.
What to watch
What to watch: how much of the $500 billion in MOUs actually converts into signed, funded facilities in the next two quarters, and whether credit rating agencies extend investment-grade treatment to GPU-backed debt more broadly the way they did for CoreWeave.
