Interviews

Mark Cuban warned AI infrastructure spending could trap America in debt

The investor told Diary of a CEO that hyperscalers borrow more than they earn while racing to power data centers.

Mark Cuban said the United States could slide into a debt bubble driven by artificial intelligence infrastructure, with cloud giants borrowing faster than their cash flow can support.

Cuban spoke on Diary of a CEO as capital floods into GPUs, power, and data centers. He framed the moment as bigger than prior tech cycles but warned that brute-force spending, not breakthrough efficiency, is what lenders are financing.

He said the market already feels bubbly, though not like the dot-com crash of 2000. Stocks now trade like meme assets, he argued, swinging double digits on penny-wide earnings misses while intrinsic value barely moves.

On AI itself, Cuban said models are predictive rather than smart. They do not grasp consequences or ethics, he said, and four leading chatbots will answer the same question four different ways. He still called himself a proponent who uses agents daily for Cost Plus Drugs and other work.

The debt thread is what he stressed for policy-minded listeners. Hyperscalers compete for the same pool of lenders as a rising federal deficit, he said, while interest on U.S. debt already runs about a trillion dollars a year. Meta, in his telling, borrows well beyond its cash flow; Nvidia's balance sheet effectively subsidizes much of the ecosystem.

Cuban cited Ray Dalio's view that the country may already be past the point where heavy debt erodes currency strength. He said he is playing defense, keeping cash and spreading holdings across countries while waiting, in Warren Buffett's style, for a turn.

Until AI needs less power to do more, not more power for incremental gains, he said excess data-center capacity could end up idle, a metaphor he offered for pickleball courts rather than perpetual compute farms.

Source video: Mark Cuban: AI Could Bankrupt Us Into A Deep Debt Bubble!