BitMEX co-founder Arthur Hayes has outlined a scenario in which a slowdown in artificial intelligence spending could eventually create a new wave of dollar liquidity, potentially benefiting Bitcoin and other cryptocurrencies. Hayes presented the argument in his latest essay, “Safety First,” published September 21.
His thesis challenges the explanation that safety concerns alone are driving some U.S. AI companies to slow the pace of frontier model development. Hayes argues that weaker demand for expensive AI services could be an economic factor, particularly as competition from cheaper models puts pressure on the economics of training and inference.
Anthropic CEO Dario Amodei has publicly called for the industry to slow the development of increasingly capable AI systems so safety measures can catch up. OpenAI executives have also supported greater attention to AI safety. Hayes disputes the motivation behind the slowdown, presenting his interpretation as an economic argument rather than an established explanation for the companies’ decisions.
Hayes Sees AI Infrastructure Debt as the Pressure Point
Hayes argues that the AI boom has created a large financing structure dependent on continued demand for computing capacity. He estimates that more than $1 trillion of investment-grade debt and hundreds of billions of dollars in lower-quality loans are connected to the infrastructure supporting AI development.
The concern is that data centers and semiconductor purchases were financed on assumptions about continued AI compute demand. If companies reduce training expenditures and become more efficient with inference, Hayes argues that revenue supporting those investments could fall while the underlying debt remains outstanding.
Recent financial reporting shows that AI infrastructure financing has become increasingly complex. The Financial Times reported that technology companies have provided as much as $300 billion in residual-value guarantees connected to AI data centers and chips, allowing some infrastructure debt to sit outside the companies’ direct balance sheets.
That does not establish that the AI credit market is heading toward a crisis. It does, however, demonstrate why a sustained change in AI spending could matter beyond technology companies themselves. Creditors, infrastructure owners, semiconductor suppliers and financial institutions could all be affected if expected cash flows fail to materialize.
Hayes’s argument then moves from corporate debt into the insurance industry. He cites research suggesting that insurers could have significant indirect exposure to AI-related private credit and reinsurance structures. If those assets were marked down sharply, he argues, losses could eventually create pressure for government intervention.
Bitcoin Becomes the Liquidity Trade
Hayes sees two potential government responses if his scenario unfolds. Washington could become a “compute buyer of last resort,” supporting demand for AI infrastructure through government purchases or long-term commitments. Alternatively, authorities could provide financial support to insurers exposed to losses from AI-related debt.
Neither response has been announced as a policy for an AI credit downturn. Hayes is presenting a hypothetical path in which a financial shock creates pressure for policymakers to expand government spending, credit support or monetary liquidity.
That distinction is important because the Bitcoin argument depends on the second stage of Hayes’s thesis. An AI slowdown by itself would not automatically increase Bitcoin liquidity. His expected outcome requires a policy response that expands the supply or circulation of dollars through government spending, financial support or related measures.
Hayes argues that such an environment would favor scarce assets, particularly Bitcoin. His reasoning is that increased dollar liquidity can push investors toward assets viewed as protection against monetary expansion, potentially lifting Bitcoin and other cryptocurrencies.
The thesis comes as AI financing continues to attract scrutiny. Apollo has estimated that AI-related borrowing could support more than $2 trillion of additional investment-grade debt through the decade, while the Financial Times has documented increasingly large guarantees supporting AI infrastructure financing. These figures highlight the scale of the investment cycle but do not prove Hayes’s predicted credit crisis.
For crypto markets, Hayes’s argument therefore represents a liquidity scenario rather than a direct prediction that an AI crash will occur. The critical variables are future AI demand, the performance of data-center debt, the exposure of financial institutions and the response of U.S. policymakers if credit conditions deteriorate.
Hayes ultimately sees both potential government responses leading toward the same destination: greater monetary liquidity. If that scenario materializes, he believes Bitcoin could benefit as investors seek assets with limited supply. Whether it happens will depend on developments across AI demand, credit markets, insurance exposure and U.S. fiscal and monetary policy, making the thesis a conditional macroeconomic argument rather than a guaranteed Bitcoin outcome.
About The Author
