Nvidia CEO Jensen Huang speaks to media members outside a restaurant in the Hongdae district of Seoul, South Korea, on June 5, 2026.
Cho Sung Joon | Bloomberg | Getty Images
Jensen Huang built the world’s most valuable company by pioneering the specialized computer chips behind the artificial intelligence boom.
To keep their vision of the future within reach, NVIDIA’s founders are now trying a different kind of engineering. The challenge is to convince Wall Street investors that these chips are long-term financial assets, akin to commercial real estate or toll roads.
His bet is on outpacing China’s AI development.
This week, Nvidia announced deals with six of the world’s largest asset managers. black rock, black stone, Apollo, K.K.R. brookfield and Goldman Sachs. The goal was to assemble a $500 billion pipeline to finance the construction of data centers and GPU clusters for companies that lack the credit ratings or the cash to buy millions of dollars of silicon.
One key premise is key to his plan, which Huang announced on a CNBC segment featuring leaders from all six Wall Street companies. The idea is that Nvidia’s graphics processing units will retain their value over time and behave more like traditional hard assets rather than rapidly depreciating consumer electronics.
“Nvidia’s AI Factory platform is truly an investable asset, an infrastructure asset,” Huang said. “That’s because it’s highly productive, revenue-generating, fungible, used by nearly every cloud service provider, and capable of running any AI model.”
In standard asset-backed financing, banks lend money. If a borrower defaults, banks can seize and sell assets such as buildings, warehouses, and cargo ships to get their money back. These physical assets have established secondary markets and can last for decades.
However, the lifespan of cutting-edge GPUs is not yet determined.
New chips will enhance the training of frontier models, but after a few years they will be relegated to low-margin inference work, changes that will directly impact resale and collateral values.
“The primary risk here is depreciation,” said Ben Emmons, founder of FedWatch Advisors. He made similar asset-backed loans for IndyMac before joining Pimco as a portfolio manager. Nvidia chips “could drop in value sooner than expected,” he said.
High yield?
In particular, Emmons said he believes the biggest threat to Nvidia’s funding model comes from China, which is rapidly increasing its domestic computing capacity and may choose to flood the market with lower-cost silicon in a price war.
If Chinese production causes hardware prices to plummet, the collateral backing hundreds of billions of dollars in private loans could wear down much faster than the terms of the debt itself, exposing investors to losses, Emmons said.
To at least partially compensate for that risk, Emmons estimates that investors will treat GPUs as highly depreciable equipment rather than real estate, demanding high yields in the 11% to 17% range depending on their place in the capital structure.
Moreover, the Bank of America Securities note said the borrowers are likely to be non-investment-grade companies excluded from traditional debt markets, such as AI startups and neo-clouds.
If these high-risk borrowers fail, Wall Street fund managers will be forced to seize used chips and resell them on potentially declining markets.
Whatever risks China poses will not materialize soon. Huawei, China’s leading provider of AI chips, has been on the U.S. Department of Commerce’s Entity List since 2019. And in May, the U.S. government announced that Huawei’s Ascend AI chip violated U.S. export controls, preventing U.S. companies from using the chip.
Meanwhile, Nvidia remains the leading supplier of AI chips in the US, with more than 75% market share by most estimates.
And for now, the economic climate remains in Mr. Hwang’s favor. As hyperscalers race to increase capacity, Huang noted that the shortage has pushed rental prices for Nvidia’s H100 chips up from about $1.70 per GPU hour in late 2025 to about $2.35 per GPU hour this year.
Importantly, NVIDIA claims that its CUDA software layer, which allows developers to run AI workloads on its GPUs, can continuously improve hardware performance after deployment, keep older chips productive, and generate revenue for longer than traditional accounting models predict.
The future of AI advancements and hundreds of billions of dollars of investor money may depend on who gets it right.
—CNBC’s Ari Levy contributed to this report.
