Wall Street bets $500b on Nvidia AI chips as long-term assets

Nvidia, the AI era's front-runner, and Wall Street are placing enormous bets on the long-term value of AI chips — wagering that Nvidia's chips can defy the industry's traditional pattern of rapid depreciation and establish themselves as a durable new asset class.

The Financial Times reported Wednesday that Wall Street investors participating in Nvidia's plan to build a $500 billion (about 710 trillion won) AI infrastructure financing platform are betting that the value of Nvidia's AI chips will not fall sharply over time.

Nvidia announced Monday that it had signed MOUs with six financial firms — Apollo Global, BlackRock, Blackstone, Brookfield Asset Management, Goldman Sachs and KKR — to create large dedicated funding pools for AI infrastructure development.

Executives at the participating firms told the Financial Times that the initiative is based on expectations that Nvidia chip prices will remain elevated longer than originally anticipated, as competition to secure the key components driving the AI boom continues.

According to those executives, Nvidia CEO Jensen Huang sees the deal as laying the groundwork for a new asset class built around AI chips — one that could become a suitable investment for the $22 trillion private equity market.

Investment heavyweights including Blackstone's Jon Gray and BlackRock's Larry Fink have also expressed willingness to deploy capital into data centers needed to train and run the latest AI models.

The financing model carries risks, however. A central question is whether investors have correctly assessed the future demand and value of Nvidia chips, which serve as the core collateral.

Ben Bajarin, a technology analyst at Silicon Valley-based consulting firm Creative Strategies, said the entire proposition hinges on whether sustained investment remains viable. "There could be overproduction, a drop in demand, or improvements in models that reduce the need for so much computing," he said.

Rental cars and commercial aircraft retain physical value even when a customer defaults, making it relatively easy to find new lessees. AI chips, by contrast, face uncertain long-term value because technology advances rapidly and demand can shift.

In chip-lease-backed debt, lenders often require full repayment within three to five years out of concern that the underlying assets could become nearly worthless. In that scenario, the key question becomes how much cash-generating power Nvidia hardware can demonstrate over that period.

Huang has pushed back on those concerns, arguing that explosive demand for computing power has kept even Nvidia's older H100 chips valuable longer than expected. He has pointed to rising chip rental rates and noted that the six-year-old A100 chip is still in use well beyond its originally projected lifespan.

One private equity executive participating in Nvidia's financing platform told the Financial Times that GPUs — Nvidia's AI computing chips — have intrinsic value. "Nvidia has more than 10 years of experience leasing GPUs," the executive said.

Another fund executive said private equity plays a unique role in this space, predicting that GPU leasing will standardize pricing and improve efficiency for AI groups.