AI Debt Bubble Unwinds: Bitcoin Could Soar to $1 Million Amidst Crisis

Arthur Hayes, co-founder of BitMEX and Maelstrom chief investment officer, has warned that $1.5 trillion in AI-related debt poses a significant risk to the financial system. Since late 2022, this debt has absorbed nearly all of the U.S. M2 expansion over the same period, starving Bitcoin of liquidity and setting up a credit crisis that could dwarf the 2008 subprime collapse.

Hayes argues that the AI credit cycle is structurally similar to the 19th-century railroad build-out in its eventual failure mode. He specifically flagged GPU loan mismatches, where financing is amortized over five to six years while leading AI hardware becomes functionally obsolete for frontier workloads in roughly two years.

Hayes also noted that competitive pricing pressure from Chinese AI models could force U.S.-built AI services to match their pricing, causing the cash flow assumptions underpinning those GPU loans to deteriorate rapidly. This dynamic is not purely speculative, as a Bank for International Settlements (BIS) bulletin on AI-related private credit documented that this asset class has grown from near zero to over $200 billion, representing nearly 8% of total private credit.

When the AI credit cycle turns, Hayes believes central banks and fiscal authorities will inject liquidity to stabilize the banking system. However, he argues that once investors have experienced AI-related losses at scale, new capital will not rotate back into AI because the sector will no longer meet its cost of capital. Instead, this capital 'goes straight to crypto,' implying Bitcoin absorbs a disproportionate share of crisis-era money printing precisely because it sits outside the institutions and asset classes damaged by the unwind.