BitMEX co-founder Arthur Hayes has reframed the AI safety debate as a credit story with significant implications for Bitcoin. Hayes argues that warnings from AI labs like Anthropic and OpenAI to slow development may actually signal slack demand at current prices rather than existential risk. He notes that cheaper AI models from Chinese rivals, priced at roughly one-hundredth of US offerings, are pressuring the expensive training race. Hayes warns that a potential slowdown in AI infrastructure spending, backed by over $1 trillion in debt, could trigger a financial crisis more akin to 2008 than 2000. This scenario, he predicts, would force government money-printing that historically drives Bitcoin prices higher.
Arthur Hayes has published a new essay, “Safety First,” arguing that recent calls from US AI labs to slow development may be a credit story rather than a safety debate. Hayes notes that rivals offer competing models at a fraction of the cost, suggesting the warnings signal slack demand, not existential risk.
Hayes models the three dominant AI labs as still financially unviable on a full-cost basis. Their forecast compute requirements generate funding needs weighing over a trillion dollars of investment-grade debt and hundreds of billions in lower-rated obligations.
These funds support data centers, power arrangements, and chip supply chains. Hayes sees the risk as 2008 rather than 2000, warning that a five-to-seven-year build-out loan book could suffer if training decelerates.
Hayes outlines two policy responses, both requiring money creation. Washington could become a last-resort compute buyer for national security, or it could backstop insurers caught with de-rated AI debt.
Using Bitcoin‘s price of approximately $65,000 as a comparison, Hayes suggests the cryptocurrency is highly sensitive to the resulting liquidity. Similar circumstances were observed in 2020 and 2023 when Federal Reserve interventions coincided with rallies.
The thesis has three trigger points: S-1 filings from Anthropic and OpenAI revealing unit economics, rating actions on data center debt, and Treasury or Fed comments on AI infrastructure. Hayes suggests the outcome is asymmetric for crypto, as bad debt socialization typically expands the money supply, historically buoying Bitcoin.
