Hayes: Debt-fueled AI boom could spark bailouts, lift Bitcoin

Arthur Hayes, co-founder of BitMEX, wrote in a Tuesday blog post that heavy borrowing to fund artificial intelligence data centers and power infrastructure could trigger a credit crisis similar to 2008.

Hayes argued investors have treated large-scale spending on data centers and energy as high-growth technology investment when the projects resemble leveraged real estate. He warned lenders may finance extensive construction and that a slowdown in AI capital expenditure could reveal weaker borrowers. He described the situation as a “credit story like 2008 and not an earnings story like 2000.”

In Hayes’s scenario, fiscal and monetary authorities would respond to any credit shock with broad liquidity measures and targeted support that could flow into risk assets, including Bitcoin. He outlined a potential price path in which Bitcoin trades in a $60,000 to $70,000 range, with possible downside near $50,000, before a liquidity-driven recovery lifts the token above $1 million. He also forecast Ether reaching $5,000 by year-end and said his investment vehicle, Maelstrom, plans to accumulate a significant ETH position while selling out-of-the-money ETH put options.

Hayes has previously connected AI competition and capital flows to crypto markets. In May he linked U.S.-China AI rivalry to increased bank lending and fiat creation that could benefit Bitcoin. In early June he reduced exposure to tokens HYPE and NEAR, citing the risk that large AI-related public listings could draw capital away from crypto.

Corporate commitments for AI infrastructure are large. Major technology companies have committed about $1.09 trillion to future data-center leases that have not yet started. That figure reflects undiscounted payments spread over many years and is not equivalent to immediate debt. By contrast, companies report roughly $285 billion in lease liabilities already on their balance sheets.

The financial impact of those commitments varies across firms. One large cloud and enterprise software company has a debt-to-EBITDA ratio near 4.3, while several others report ratios below one. An analyst at S&P Global noted that some data-center leases run 15 to 19 years, while customer contracts typically last no more than five years, creating a potential duration mismatch between long-term lease obligations and shorter-term revenue contracts.

Hayes framed his view as a chain of events-heavy lending, a credit shock, policy responses and a large crypto rally. He and other observers acknowledge that the scenario is speculative and depends on how quickly AI capital expenditure slows, which borrowers are weakest, and how policymakers respond to any financial stress.