Jeff Gundlach says making AI chips an asset class looks like a market top

Jeff Gundlach has criticised Nvidia’s plan with six asset managers to mobilise more than $500bn for AI infrastructure, asking why assets of unknown life are collateral for long-term debt. Mark Cuban compared chips as an asset class to crypto.

Jeff Gundlach thinks Wall Street has told on itself. “Assets of unknown life as collateral for long term debt?” the DoubleLine chief executive wrote on X, saying Nvidia’s new financing consortium “will not likely age well.”

He is talking about a large arrangement. Nvidia signed agreements last week with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR to build financing platforms intended to mobilise more than $500bn so customers can buy access to its chips.

His analogy is deliberately undignified. Gundlach compared it to issuing 30-year asset-backed securities against warehouses of bananas, even “newly engineered bananas of unknown life.”

Underneath the joke is a duration question. The debt is long-dated, and the collateral is a class of hardware the industry replaces at speed.

The participants have said as much themselves. KKR’s co-chief executives described compute as critical infrastructure requiring “long-duration capital,” while Jensen Huang argues that continuous CUDA software work is “extending its useful life and improving its economics over time.”

Those are not the same claim. One says the money must be patient, the other says the asset will keep up, and the second is doing considerably more work.

Gundlach’s wider point is about pattern recognition. Nobody rings a bell at the top of risk markets, he wrote, but investors should watch for proclamations of new asset classes built on financial engineering and helped along by what he called “questionable” credit ratings.

Others are making versions of the same argument. Mark Cuban wrote that “chips as an asset class will be the new crypto,” and Michael Burry has been shorting AI companies while arguing that today’s processors will be obsolete before the spending is repaid.

The credit market has already reacted once. Nvidia’s own credit default swaps hit a record after it announced $750bn of AI deals, which is the market pricing the same worry rather than tweeting about it.

The systemic version of this is already on the table, with the Bank for International Settlements warning that an AI bust could hit credit markets as hard as 2008. Gundlach’s objection is narrower and more answerable than that. It is not whether AI works, but whether a graphics processor is still worth lending against in year seven.