Are Nvidia AI factories 'an investable asset class'?

Nvidia and private equity and investment firms have signed a $500 billion memorandum of understanding to fund AI infrastructure. The big question is whether Nvidia's argument that AI factories are "an investable asset class" holds up to scrutiny.

First, the news. Nvidia said Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR will create funding platforms for AI infrastructure. The deal may happen or not. This announcement falls into the same category as Stargate where there are big sums of money and a memorandum of understanding and no follow-through. All you need to know about this AI infrastructure funding deal is at the bottom of the release: "These partnerships remain subject to execution of the final agreements."

There was a lot of pomp and circumstance on CNBC for this financing arrangement, but the real sell was the pitch that Nvidia AI factories are investments in durable infrastructure. Note that Nvidia Jensen Huang wasn't talking about AI data centers in general and he sure wasn't talking up AMD or any rival. He was just talking about Nvidia's stack being an asset class.

"We began by building chips; today, we are helping create a new class of productive, investable infrastructure: AI factories. In AI, compute is revenue," said Huang.

Well Jensen, compute is a cost. Compute is revenue for cloud hyperscalers and a few others. For the rest of us, compute is a cost.

• Enterprises need better AI value metrics and vendors need better pricing models
• Enterprise AI grows up, vendors are noticing
• Here’s what we learned about AI projects from enterprise buyers so far
• What if our AI infrastructure is already built?

In a post on X, Huang moved his argument a bit more. He noted that Nvidia compute prices for older GPUs are holding up and increasing for newer platforms. Huang said:

"Nvidia AI factories are different. Their value is not fixed at installation: CUDA continuously improves their output; the installed base remains productive well beyond its initial depreciation period; and the same standard architecture serves a deep, growing global market of AI workloads.

These are the characteristics of an investable infrastructure asset: it produces revenue, serves a broad market, improves in performance over time and can be redeployed."

Huang went on to argue that the AI funding isn't circular financing. He added that Nvidia is providing a backstop for 25% on a "project-by-project basis". "That support is limited, residual-value based and designed to complement — not replace — independent underwriting," said Huang.

Give Huang credit for a move that may roughly create Nvidia Capital, which may wind up rhyming with GE Capital (it worked well until it didn't). The bigger question is whether you should buy into this investable asset class argument.

A few thoughts:

• This move sets up an asset-repackaging go round that can offload data center projects to various investors. Think mortgage-backed securities.
• The land and buildings in the data center ecosystem are an asset. The systems inside those facilities depreciate like a used car.
• If you have any doubts that the AI bubble (if there is one) is really about commercial real estate. Data centers are a real estate asset class, but face rapid obsolescence, power constraints and a lot of upfront capital.
• CBRE released its mid-year review of the commercial real estate sector and noted that data center demand is strong, but there are some nuances. "Counterparty and tenant risk remain a focus in 2026 and differentiate landlord strategies. While certain landlords and operators are willing to lease significant capacity to non-investment-grade tenants, others are opting to only serve the world’s largest technology companies. In the event of a slowdown, the bifurcation of financial health of operators will reveal itself," said CBRE.
• Nvidia is trying to line up funding because investors are more skeptical of firms like Blue Owl, off-balance sheet deals, special purpose vehicles and debt loads of tech giants that are often understated due to creative financing arrangements.
• Time will tell if Nvidia's stack is an investable asset class. If Bitcoin is investable you can make the argument that AI factories are too. However, Nvidia's take that only its stack is investable is simply Huang talking his game. This whole asset class argument for just Nvidia is accounting fantasy.
• If this financing setup actually happens (call me a MOU skeptic) and works the Huang will look like a GPU and financing genius. If this merry band of financiers fails, Huang will have rang the bell at the top of the AI bubble.