Wu Xie learned that in his latest article (Safety First), Arthur Hayes said he believes that U.S. AI companies such as Anthropic and OpenAI emphasizing “safety first” and slowing AGI development are not primarily doing so for safety reasons, but because there is insufficient AI demand at current prices. If training spending declines, it will directly weaken data center and chip computing demand. Hayes said that although the AI labs themselves are not yet profitable, their computing needs have already supported more than $1 trillion in investment-grade debt and tens of billions in sub-investment-grade debt and loans. Hayes further believes that if AI compute demand falls short of expectations, related debt could be devalued, and risks could be transmitted through an insurance and reinsurance system controlled via private equity. The U.S. government may ultimately choose to become the “last compute buyer,” or avoid credit events by rescuing the insurance industry. No matter which path, he believes the end result will be more money injection, thereby pushing up the prices of Bitcoin and some other crypto assets.

Wu Shuo reminds that Arthur Hayes' market forecasting record has lower accuracy and his views often change; it is recommended to focus on his analytical framework rather than his specific price and time predictions.