Arthur Hayes warns that around $1.5 trillion in AI-driven debt could trigger a financial crisis larger than the 2008 meltdown, ultimately forcing massive money printing that may propel Bitcoin toward $1 million per coin.
Speaking on the Bankless podcast on June 22, 2026, the BitMEX co-founder and Maelstrom CIO said AI-related borrowing since late 2022 has absorbed nearly all growth in the U.S. M2 money supply. In his view, this has diverted liquidity away from Bitcoin while quietly building systemic risk. If the cycle reverses, Hayes expects a credit collapse that could exceed the scale of the subprime crisis, with Bitcoin emerging as a major beneficiary.
His thesis is not just a price prediction but a broader argument about large-scale capital misallocation. Hayes believes the financial system is facing an unprecedented imbalance, where excessive investment in AI infrastructure will eventually unwind—redirecting capital into alternative assets like Bitcoin.
He explained that funds which might have flowed into crypto have instead been poured into data centers and GPU clusters financed with long-term debt. Hayes likened the AI boom to the 19th-century railroad expansion, highlighting similar risks of overinvestment. A major weakness, he noted, is the mismatch between financing and technology cycles: loans are structured over five to six years, while AI hardware can become obsolete in roughly two.
Another pressure point comes from global competition. If U.S. AI firms are forced to cut prices to match cheaper Chinese models, expected revenues could fall sharply, undermining the loans backing this infrastructure. Hayes described such a repricing as a potential trigger for a major credit event—one he believes could surpass the subprime crisis.
Data from the Bank for International Settlements supports the growing scale of the risk. AI-linked private credit has expanded from near zero to over $200 billion, now accounting for about 8% of the market. At the same time, large tech firms are increasingly shifting AI-related debt off their balance sheets through special-purpose vehicles and leasing structures, creating less transparent channels for financial contagion.
In Hayes’ view, the policy response to a crisis would be predictable: central banks and governments would inject large amounts of liquidity to stabilize the system. He suggested authorities would effectively flood markets with fiat money to offset years of excessive AI investment.
The key question is where that liquidity flows. Hayes argues that after significant losses in AI, investors are unlikely to return to the sector. Instead, capital would rotate into crypto—particularly Bitcoin—because it exists outside the traditional financial system and is less exposed to the damage caused by the unwind.
A $1 million Bitcoin would imply a market value of roughly $21 trillion, requiring liquidity injections far beyond those seen during the COVID-era stimulus. Hayes acknowledged the timing is uncertain, saying the AI bubble could burst soon or take years. His thesis depends on a crisis-scale monetary response rather than gradual easing.
For this scenario to materialize, widespread defaults in AI-related credit—especially among smaller GPU lenders and leveraged data center operators—would need to trigger aggressive policy action. If institutions begin treating Bitcoin as a hedge against currency debasement, the capital rotation he describes becomes more plausible.
However, risks remain. In past crises, capital has first moved into traditional safe havens like government bonds and gold. Bitcoin, which has often traded like a risk asset during periods of stress—as seen in March 2020—could initially fall alongside AI-linked equities before benefiting later.
Hayes’ own positioning reflects caution. As of June 2026, he described himself as consistently long Bitcoin while holding significant cash in Treasury bills and reducing exposure to higher-risk tokens like NEAR and Hyperliquid. He emphasized capital preservation as key to surviving market cycles, framing the $1 million Bitcoin target as a potential cycle peak rather than a near-term expectation.
