Wall Street’s new long-term play on AI chips signals a seismic shift in infrastructure valuation

Wall Street and private capital are increasingly viewing AI hardware as enduring infrastructure, with Nvidia leading a financing revolution aimed at unlocking hundreds of billions in long-term value, even amid scepticism about hardware longevity and demand sustainability.

Wall Street is moving to treat artificial intelligence chips less like fast-depreciating electronics and more like long-lived infrastructure. Nvidia’s latest financing push, involving firms including Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR, centres on the idea that the GPUs powering AI systems can retain meaningful value well beyond the timeframes usually associated with computing hardware. Reuters reported that the structure could ultimately mobilise about $500 billion for AI build-outs, a scale that signals how deeply private capital now wants into the market for computing capacity.

The bet rests on a simple but risky premise: demand for compute will stay strong enough to support chip prices, rental income and resale values for years. Nvidia chief executive Jensen Huang has argued that even older chips remain useful because AI developers, cloud providers and enterprises still need vast amounts of processing power for training and inference, the latter being the work of answering prompts and running models. In practice, that means premium chips such as Nvidia’s GPUs may be deployed in high-end research jobs while older units continue to serve less demanding tasks. Axios and Tom’s Hardware reported that the financing plan is designed to make AI infrastructure easier to fund by turning it into an asset class that can attract institutional money.

That logic has drawn support from some of the biggest names in private capital, but it is not without sceptics. Ben Bajarin of Creative Strategies warned the Financial Times that the whole model depends on continued investment and that there is a real risk of overbuilding, slower demand or more efficient models that need less computing power. The concern is familiar to anyone who has watched technology assets age quickly: unlike aircraft or cars, chips do not have a deep, predictable second-hand market if the underlying economics weaken. Some lenders are therefore limiting repayment periods to three to five years, reflecting the assumption that the hardware may be worth far less after that point.

Nvidia is trying to reduce that risk by giving investors a floor under the collateral. The Financial Times reported that the company is guaranteeing at least 25% of value over the life of some lease agreements, effectively taking the first loss if chip prices fall more sharply than expected. Executives involved in the deal said that this kind of backstop could help package GPU leases into securities that would be saleable to insurance companies and other large buyers of debt. In that sense, the transaction echoes the kind of structured finance more commonly associated with private equity buy-outs than with semiconductor sales.

For Nvidia, the arrangement also offers a more practical benefit: it may reduce pressure on the company to fund customers directly. BofA Securities analyst Vivek Arya has said the backing of large financial institutions could allow Nvidia to scale back supplier financing, under which the chipmaker itself helps customers raise money in capital markets. That would shift more of the risk onto the broader consortium rather than Nvidia’s balance sheet. Even so, the model depends on a judgement that remains unproven at such scale: that AI computing power will stay scarce and valuable long enough for Wall Street to earn a return on hardware that is already being redesigned, redeployed and replaced at speed.