Arthur Hayes in his new essay 《Safety First》 delivers a hard conclusion: the U.S. insurance industry is effectively already insolvent.
His reasoning: most AI large-model companies don’t make money, yet they support more than $1 trillion in debt; analysts estimate that behind U.S. insurance companies there is $1.54 trillion in affiliated reinsurance, while the industry’s total surplus is only $657 billion—remove that piece, and among the top 30 insurers, 29 are technically bankrupt. Once the credit rating of the AI data-center debt is downgraded, they can only wait for state guaranty funds to step in, while the payout cap on any single policy is just $250,000–$300,000.
My take: his “either rescue computing power or rescue insurance—ultimately it comes down to printing money, and $BTC benefits” chain-of-logic is plausible, but the third-party calculations he cites are from private audits, not confirmed by regulators—so don’t treat them as established fact. What really needs scrutiny is the rating actions on that batch of AI data-center debt.
Do you think the first to blow up will be the insurance companies, or the AI capital expenditures?
His reasoning: most AI large-model companies don’t make money, yet they support more than $1 trillion in debt; analysts estimate that behind U.S. insurance companies there is $1.54 trillion in affiliated reinsurance, while the industry’s total surplus is only $657 billion—remove that piece, and among the top 30 insurers, 29 are technically bankrupt. Once the credit rating of the AI data-center debt is downgraded, they can only wait for state guaranty funds to step in, while the payout cap on any single policy is just $250,000–$300,000.
My take: his “either rescue computing power or rescue insurance—ultimately it comes down to printing money, and $BTC benefits” chain-of-logic is plausible, but the third-party calculations he cites are from private audits, not confirmed by regulators—so don’t treat them as established fact. What really needs scrutiny is the rating actions on that batch of AI data-center debt.
Do you think the first to blow up will be the insurance companies, or the AI capital expenditures?