BitMEX co-founder Arthur Hayes has reframed the AI safety debate as a credit story with clear Bitcoin implications. In his latest essay, Safety First, published on Crypto Trader Digest on Substack, Hayes writes that recent warnings from US AI labs, including Anthropic and OpenAI, that AI should be slowed may signal slack demand at current prices and not death.
Why does no one buy AI? Because no one is prepared to pay for it at US prices. Other models developed by Chinese rivals are priced at around a hundredth as much, raising the stakes of the compute-intensive training race.
If labs scale back training budgets and focus on efficiency, demand for data centers, chips from Nvidia and Broadcom, and cloud capacity from Microsoft and Google may all soften. On crypto, the policy response is the key.
The $1 Trillion Debt Behind AI Infrastructure
Arthur Hayes focuses on the financing supporting the build-out. He models the three dominant AI labs still financially unviable on a full-cost basis, but their forecast compute requirements generate funding requirements outweighing a trillion dollars of investment-grade debt and hundreds of billions of lower-rated debt.
These fund data centers, power arrangements, and chip supply chains, often supported by hyperscaler balance sheets. He sets the risk as 2008 rather than 2000. Data centers are business real estate with rapidly-depreciating chips rapidly depreciate as fresher models run at higher performance per watt.
A five- to seven-year build-out loan book could suffer shortages if the training pace decelerates. Downgrades would force insurers and private credit funds to over-allocate. An outright slump would impact risk assets first.
Also Read: Arthur Hayes: Why AI Bubble Could Trigger 2028 Crisis
Why Money Printing Would Drive Up Bitcoin
Arthur Hayes lays out two paths for government, both involving more dollar liquidity. A situation where Washington becomes a last-resort compute buyer to take on excess capacity to help with national security.
Hoping to backstop insurers that could be caught with de-rated AI debt on their books, he writes, analyzing how private equity firms are acquiring insurers and placing AI debt into their holding company’s general accounts, funded by related reinsurance vehicles with small capital.
Both options require money creation on the Fed balance sheet or taxpayer backing of the small reinsurance vehicles. Using Bitcoin’s $65k on CoinMarketCap as a comparison, effects on its price are quite sensitive to liquidity, along with Ethereum, Solana, USDT and USDC stablecoins, and DeFi protocols like Aave tracking real rates, and inflows to Bitcoin ETFs managed by BlackRock and Fidelity.
Also Read: Arthur Hayes Targets $126,000 Bitcoin Price as US Debt Tops $40 Trillion
Expand context and implications
The thesis builds on an argument that views Bitcoin as indicative of fiat-printed liquidity, as similar circumstances were seen in 2020 and 2023 when the Fed intervened, and the Treasury reduced its balance, both times managed to coincide with a rally.
Data from Glassnode and DefiLlama indicates stablecoin supply growth is perennially transitory, sends signals for Coinbase/Binance exchange flows, and has and will continue to impact funds betting on Bitcoin as a macro-hedge, custodian collateralization, developers on ethereum layer-2s, and regulators, including the SEC and state commissioners watching private credit.
Source: Bloomberg.com
The thesis has three trigger points. First, S-1s of Anthropic and OpenAI will reveal unit economics, including cost per token. Second, rating actions on data center debt will provide stress signals. Third, Treasury or Fed comments on AI infrastructure or insurance funds will verify the channel.
Also Read: Ethena Price Eyes Breakout to $0.3613 as Arthur Hayes Expands ENA Holdings
Bitcoin’s Asymmetric Upside
Arthur Hayes suggests that the outcome is asymmetric for crypto, in the sense that it can be positive if capacity is taken on the books or bad debt is socialized, in which case money supply expands.
Source: Bombay Chamber
The risk is that support is delayed or bundled with an even tighter leverage regime, muting the transmission. But, by anchoring the safety rhetoric to debt sustainability, he offers a scenario where an AI bust results not in austerity but in market conditions that historically buoy Bitcoin.
Also Read: Arthur Hayes Says Trump Needs “AI President” Role 2026
Be the first to comment