Jensen Huang’s $500 billion Nvidia bet: Can AI chips become Wall Street’s next big asset?

This week, Nvidia announced agreements with six of the world’s biggest asset managers — BlackRock, Blackstone, Apollo, KKR, Brookfield and Goldman Sachs. The firms are expected to help create a financing pipeline of as much as $500 billion for data centers and GPU clusters.

Many companies want to build AI infrastructure but do not have enough cash or a strong enough credit rating to buy millions of dollars worth of Nvidia chips themselves. Wall Street financing could allow them to build those facilities and pay for the chips over time.

For Huang, the plan depends on one important question, will Nvidia’s chips remain useful and valuable for long enough to make this financing model work?

Why Nvidia wants investors to back its chips

Huang wants investors to look at Nvidia’s AI infrastructure differently from ordinary electronic equipment. In traditional asset-backed financing, banks are willing to lend money against physical assets because those assets can be taken back and sold if the borrower fails to repay the loan. Buildings, warehouses and ships, for example, can remain useful for many years and have established markets where they can be resold.

Huang believes Nvidia’s AI systems can work in a similar way because they generate revenue and can be used by many different companies.

“Nvidia’s AI factory platform is really an investable asset, an infrastructure asset,” Huang said. “The reason for that is because it’s productive, it’s revenue generating, it is fungible, it’s used by just about every cloud service provider, it runs every AI model.”

If Nvidia’s GPUs continue to generate enough income for several years, investors can lend money against them with greater confidence. But there is a major difference between an AI chip and a building or a road. Technology changes quickly.

The biggest problem is depreciation

The biggest uncertainty for Huang’s plan is how quickly Nvidia’s chips lose their value. The newest GPUs are extremely important for companies developing and training advanced AI models. But as newer and more powerful chips arrive, older GPUs can become less attractive for the most demanding workloads. They may then move into less profitable uses, such as running AI models rather than training them.

That could reduce the amount of money an older chip can generate and, in turn, reduce its value as collateral.

“Depreciation is the one key risk here,” said Ben Emons, founder of FedWatch Advisors, who structured similar asset-backed loans for IndyMac before joining Pimco as a portfolio manager told CNBC. Nvidia chips “could depreciate faster than expected,” he said.

This is important because Wall Street investors are not simply being asked to finance Nvidia’s growth. They are being asked to finance companies that may not have strong credit and to use expensive AI equipment as security for those loans.

If the equipment loses value quickly, investors could be left with assets worth much less than the money they originally lent.
China could make the risk bigger

China is another major part of the equation

China is working to increase its own AI computing capacity and develop domestic alternatives to Nvidia’s chips. If Chinese companies eventually produce large amounts of cheaper AI hardware, the global market could face pressure on prices.

That could create a serious problem for Nvidia’s financing model. If the value of AI chips falls sharply, the GPUs being used as collateral for hundreds of billions of dollars in loans could also lose value. Investors could then find themselves holding equipment that is worth considerably less than the debt it was meant to secure.

Emons believes this is one of the biggest risks facing the model. He expects investors to demand higher returns because GPUs may behave more like equipment that depreciates quickly than like buildings or other long-term infrastructure. He estimates that investors could demand yields of between 11% and 17%, depending on their position in the financing structure.

The borrowers themselves could also add to the risk. According to a Bank of America Securities note, some of the companies likely to use this financing could be AI startups and so-called neocloud companies that do not have investment-grade credit ratings. That means the risk exists on both sides. The chips could lose value, while the companies using them could also struggle to repay their loans.

Nvidia still has a strong position

The risks are significant, but they are not necessarily immediate.

Huawei, the dominant provider of Chinese AI chips, has been on the U.S. Commerce Department’s Entity List since 2019. In May, the U.S. government also said Huawei’s Ascend AI chips violate U.S. export controls, preventing American companies from using them. Nvidia continues to dominate the U.S. AI chip market. Most estimates put its share at more than 75%.

The demand for its chips also remains strong because major cloud companies are racing to build more computing capacity for AI.

Huang pointed to the economics of Nvidia’s H100 chips as evidence of this demand. Rental prices for the chips rose from around $1.70 per GPU-hour in late 2025 to about $2.35 per GPU-hour this year. For Nvidia, that rising demand supports the argument that its chips are not simply expensive pieces of electronics. They are productive assets capable of generating revenue.

Huang is also counting on software

There is another part of Nvidia’s argument that could help protect the value of its chips: software.

Nvidia’s CUDA software platform allows developers to run AI workloads on its GPUs. The company argues that regular improvements to this software can make existing hardware more useful even after newer chips enter the market. That could help older Nvidia GPUs continue generating revenue for longer than traditional accounting models might suggest.

This is important to Huang’s larger plan. If a chip can remain productive for several years, its value as an asset becomes easier to defend. A company that owns the chip can continue earning money from it, while an investor financing that chip has a better chance of recovering the money if something goes wrong. The question, however, is how long that advantage will last.

The $500 billion question

Huang’s plan is ultimately a bet on the future of AI infrastructure. Nvidia is trying to bring Wall Street deeper into the AI buildout by turning its chips into the foundation for a new form of asset-backed financing. The six major investment firms involved in the plan could help provide up to $500 billion for the construction of data centers and GPU clusters.

For Nvidia, this could create a powerful cycle. More financing would allow more companies to buy or rent computing capacity. More computing capacity would allow more AI companies to expand. And greater demand for AI computing would create more demand for Nvidia’s chips. But the entire model depends on the chips continuing to make money.

If Nvidia’s GPUs remain useful and generate strong revenue even after newer generations arrive, Huang’s argument that they are infrastructure assets could become much easier for Wall Street to accept. If the chips depreciate faster than expected, or if China brings large amounts of cheaper hardware into the market, the economics could change quickly.

That leaves one central question hanging over Huang’s $500 billion plan: how long will Nvidia’s chips remain productive enough to generate the revenue needed to support their value? The answer could determine not only how Wall Street finances the next stage of the AI boom, but also how much of Nvidia’s current dominance can survive the rapid pace of change in artificial intelligence.