律动BlockBeats
律动BlockBeats|Sep 22, 2026 01:32
Arthur Hayes: AI's' security first 'approach is actually the destruction of computing power demand, and all choices made by the US government ultimately lead to printing money, ultimately benefiting Bitcoin BlockBeats News: On September 22nd, Arthur Hayes released a new long article titled "Safety First", with the core argument that Anthropic, OpenAI, and SpaceX's claim of "safety first" guiding them to slow down AGI development is not due to concerns about human welfare, but rather due to economic reality. The market does not want AI, but rather AI at "Chinese prices", which requires intelligence that is 100 times cheaper than the current price. Hayes pointed out that 'security first' is essentially the destruction of computing power demand. If the cost of training new models decreases and laboratories shift towards efficiency optimization, customers' spending on computing power will decrease. However, the three major AI laboratories do not generate any profits, and their computing power demand supports over $1 trillion in investment grade debt and hundreds of billions of dollars in low-quality debt, which relies on off balance sheet endorsements from profitable technology companies such as NVIDIA, Broadcom, Google, and Microsoft. The real acquirer is the American insurance policyholder. Hayes cited Nick Nameth's analysis to reveal a 'self insurance scam': PE giants (Apollo, KKR, Brookfield, etc.) acquire insurance companies, stuff AI data center debt and AI impacted SaaS private credit into insurance assets, and then provide false endorsements with minimal capital through affiliated self insurance reinsurance companies. Nameth estimates that the total amount of such false reinsurance assets is $1.54 trillion. Once the debt of AI data centers is downgraded by rating agencies due to insufficient computing power demand, insurance companies will be forced to increase capital, and affiliated reinsurance companies will be unable to pay, resulting in the insurance company being insolvent. Most states in the United States have insurance coverage limits of only $250000 to $300000, and surviving insurance companies only make payments to the coverage fund after the fact, which encourages all parties involved to maximize risk-taking. When AIG was rescued in 2008, TARP funds ultimately flowed to Goldman Sachs and generated record breaking bonuses, while ordinary people only received notices of default. Hayes believed that this scene would repeat itself. For cryptocurrency investors, the conclusion is a win-win situation. If the US government chooses to become the 'last buyer of computing power', it will print money in the name of national security to support unproductive economic products, driving up financial speculation and Bitcoin prices; If the government chooses to rescue insurance companies with insolvency, it also needs to print money to cover bad AI debt, increase money supply, and push up Bitcoin. Hayes specifically pointed out that the Federal Reserve unanimously voted to raise interest rates by 25 basis points last week, and RMP bond purchases stopped on August 14th. However, commercial banks have taken over and created over $100 billion in currency, and the interest rate hike has allowed banks to earn an additional $7.5 billion in excess reserve interest annually. These funds will be used to expand loans and market speculation, and overall, the net effect is still stimulating. The volatility of the cryptocurrency market after a small increase at the end of August is coming to an end, and the supply of US dollars will continue to grow, while Bitcoin and some selected altcoins will rise. Hayes also described the situation as' incredibly wonderful ', stating that the government will not allow the free market to stop building AI data centers, there will be an oversupply of computing power spot, the use of AI agents will increase, and the surge in printing will drive investors to pursue cryptocurrency assets. [Original link]
+4
Mentioned
Share To

Timeline

HotFlash

APP

X

Telegram

Facebook

Reddit

CopyLink

Hot Reads