Bitcoin could gain from AI bust, Arthur Hayes says

Arthur Hayes says an AI credit bust could push U.S.

Share

Arthur Hayes has argued that a pullback in U.S. artificial intelligence spending could eventually trigger government support for AI infrastructure or stressed insurers, creating dollar liquidity that he expects would favor Bitcoin.

Summary

• Hayes argues weaker AI compute demand could stress debt markets and eventually increase dollar liquidity.
• Federal Reserve raised rates 25 basis points last week, challenging immediate money-printing expectations from Hayes.
• Apollo estimates AI financing needs could support over $2 trillion in debt through this decade.
• NAIC says private credit transparency and valuation risks require continued monitoring by state insurance regulators.
• Hayes expects government compute purchases or insurance support to expand liquidity and benefit Bitcoin prices.

Hayes wrote in his Sept. 22 essay, Safety First, that recent calls by major U.S. AI companies to slow frontier model development may have an economic explanation alongside the safety concerns they have publicly cited. He suggested demand for expensive AI services could prove weaker than the spending assumptions supporting data centers, chips and related debt.

His interpretation is not the explanation given by the AI companies themselves. OpenAI said in August that it temporarily slowed parts of frontier model development after cybersecurity concerns and stronger internal safeguards became necessary. Anthropic CEO Dario Amodei later called for the industry to pace model development so safety controls could catch up with capabilities. Neither statement attributed the slowdown to falling customer demand.

You might also like:

Arthur Hayes unveils FLOP tokenomics and proof of inference network

Arthur Hayes sees AI debt as the pressure point

Hayes’s argument centers on the financing behind the AI infrastructure buildout. He contends that lower demand for training and inference could weaken the cash-flow assumptions supporting data centers, semiconductor purchases and private credit tied to the sector. In his words, “Safety First is by definition compute demand destruction.”

Independent credit research confirms that large amounts of financing are being directed toward AI, although it does not establish Hayes’s projected crisis. Apollo said in August that AI-related issuance accounted for nearly 40% of longer-duration investment-grade corporate bond supply. Its economists estimated that the AI ecosystem could support more than $2 trillion of additional investment-grade debt, while public markets may absorb less than $1 trillion through 2030.

Apollo separately estimated that roughly $5 trillion could be spent on AI infrastructure through 2030. Its research said businesses and consumers would need to spend around $2 trillion annually on AI services to justify that level of infrastructure investment.

Credit dependence has continued to grow. A Sept. 21 Apollo note said consensus forecasts assume operating cash flow at five major hyperscalers — Alphabet, Amazon, Meta, Microsoft and Oracle — will rise from roughly $600 billion to $2 trillion by 2030. Apollo warned that weaker cash-flow growth could lead to wider credit spreads and reduced capital expenditure.

Hayes has made a similar argument before. As previously reported, his earlier AI credit-crisis thesis linked a potential AI downturn with credit stress followed by a monetary response that he expected would favor Bitcoin. A later AI bubble and Bitcoin liquidity argument focused more directly on leveraged data-center financing.

Insurance exposure remains disputed in Hayes’s thesis

The second part of Hayes’s scenario concerns insurers and private credit. Drawing partly on research published by Nick Nemeth, Hayes argues that affiliated reinsurance structures could leave some insurers vulnerable if AI-related debt is downgraded and must be marked lower.

Nemeth estimates that affiliated reinsurance credits across the U.S. life and annuity industry total $1.54 trillion. His analysis argues that some reinsurance assets may provide less economic protection than statutory accounting suggests. The figure is Nemeth’s estimate and does not represent a finding by U.S. insurance regulators.

The NAIC does identify private credit as an area requiring continued supervision. Its July update said private credit has less liquidity, weaker price transparency and less frequent valuation than publicly traded debt. Regulators said those characteristics have prompted closer monitoring of valuation practices, underwriting standards and sector concentrations.

NAIC material on private-equity-owned insurers similarly identifies affiliated investment management and cross-border reinsurance as areas under ongoing regulatory review. The organization counted 139 private-equity-owned U.S. insurers by June 2025.

Current industry data does not establish that the U.S. insurance sector is insolvent because of AI exposure. A recent Moody’s survey reported by the Wall Street Journal estimated direct U.S. insurer exposure to data centers at up to $20 billion, while life insurers hold far larger allocations to private debt generally. Hayes’s much larger systemic-risk argument depends on indirect exposure through private credit, reinsurance and structured financing.

Bitcoin thesis depends on a future liquidity response

Hayes presents two possible government responses if AI infrastructure economics deteriorate. One would involve Washington becoming what he calls a “compute buyer of last resort”, using government spending or offtake agreements to maintain demand for AI capacity. The other would involve financial support for insurers if losses on private credit threaten policyholder claims.

No U.S. authority has announced either policy in response to an AI debt crisis. Hayes argues that both scenarios would require increased government borrowing, banking-system liquidity or direct monetary support, which he expects would raise demand for Bitcoin and other scarce financial assets.

Current Federal Reserve policy is moving in the opposite direction from immediate monetary easing. The Fed raised its target range by 25 basis points on Sept. 16 to 3.75%–4.00%, its first increase since July 2023. The vote was unanimous, and officials said inflation remained elevated.

Reserve-management purchases have paused as well. The New York Fed scheduled no reserve-management purchases for both the Aug. 14–Sept. 14 and Sept. 15–Oct. 14 operating periods, although reinvestment purchases continue. Federal Reserve officials have repeatedly said reserve-management purchases are designed to maintain ample reserves and should not be treated as quantitative easing.

Commercial-bank credit has continued growing during the year. Fed H.8 data showed seasonally adjusted bank credit rising from $19.74 trillion in July to $19.87 trillion by the week ending Sept. 9, driven mainly by loans and leases. The figures document balance-sheet growth but do not establish Hayes’s claim that banks are replacing central-bank money creation as a deliberate stimulus program.

Hayes has previously linked Treasury and banking-system liquidity with Bitcoin. His Treasury liquidity and Bitcoin bull-market call argued in August that higher Treasury buybacks and banking liquidity could support crypto prices even without conventional quantitative easing.

AI companies are still spending despite safety calls

Evidence of an immediate collapse in AI investment remains limited. OpenAI’s official material says it slowed parts of frontier development to install stronger monitoring, alignment and security controls, while continuing to invest in model research. Its GPT-6 Astra work proceeded after the company classified the model at a critical cybersecurity capability level.

Anthropic, meanwhile, is considering another model release despite Amodei’s public request for slower industry development, Reuters reported on Sept. 19. The company is weighing competitive pressure from OpenAI alongside safety reviews and preparations for a possible future IPO.

Financing activity remains substantial. Nvidia announced in August that Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR were working on independent AI-compute financing platforms intended to mobilize more than $500 billion in third-party capital over time. The amount represents planned financing capacity, not capital already deployed.

SoftBank began marketing more than $11 billion of high-yield bonds this week to help finance its OpenAI investment, according to the Financial Times. The transaction follows large bridge loans used for earlier funding commitments and provides another current example of AI investment drawing heavily on debt markets.

Meanwhile, Bitcoin traded near $85,700 early Sept. 22 after climbing more than 6% during the previous session.

Read more:

Cardano integrates x402 to bring ADA payments to AI agents