AI debt black hole detonates insurance scams, Bitcoin must win

According to Woofun AI, Arthur Hayes pointed out that Anthropic, OpenAI, and SpaceX claim to slow down general artificial intelligence (AGI) research and development on the grounds of “safety first,” in fact to cover up the harsh economic reality that it is unprofitable. The so-called “Silicon-God” (Silicon-God) narrative created by these companies is collapsing, and the real demand for the market is a Chinese pricing model where the price is only one percent of the US, rather than expensive American technology.

This mismatch between supply and demand led to a contraction in demand for computing power, which in turn triggered a huge debt crisis behind it.

An in-depth analysis of the AI giant's financial truth reveals that profit lies are facing an impact on China's pricing. As the end of the third quarter approaches, Anthropic has been slow to go public, causing the market to strongly question the veracity of its financial reports. Hayes is eager to study its upcoming S-1 prospectus to clarify the true cost of providing users with each token and whether the profitable customer base is expanding or shrinking.

However, all questions about cost structures and profit margins are overshadowed by the pretext of “safety first”. In fact, the market is in high demand for AI, but consumers are more likely to choose a Chinese model that is only one percent of the US price. When American practitioners accuse China of poor quality products or reliance on distillation models, the market did not buy it; it only sought the cheapest intelligent service. As a result, AI giants instead suspended research and development due to concerns about human safety, in fact to seek government supervision and funding to maintain their expensive pricing system.

The computing power debt crisis is spreading rapidly, forming a complex network of trillion dollars of debt and off-balance sheet guarantees. The huge demand for computing power from leading AI laboratories such as Anthropic and OpenAI supports more than $1 trillion in investment-grade debt and hundreds of billions of dollars in low-credit rating loans. Since these laboratories do not generate any profit in total, they must rely on profitable technology companies such as Nvidia (NVDA.US), Broadcom (AVGO.US), Google (GOOGL.US), and Microsoft (MSFT.US) to provide off-balance sheet guarantees for data center leases and chip purchases. Subsequent procurement of chips and hardware depends entirely on AI Labs continuing to train cutting-edge big models and handle inference requests for customers. Once “safety first” becomes the core principle, the cost of training new models will drop from a high level, and the focus of enterprises will shift to improving the efficiency of power conversion intelligence, which will directly lead to a reduction in customer computing power expenses. Essentially, “safety first” destroys the demand for computing power, thereby shaking the foundation of the entire debt chain.

The essence of prioritizing security is to destroy demand, which in turn raises the risk of debt default. If AI capital expenses are funded from operating cash flow, the risk is still manageable, but the reality is that trillions of dollars of debt still exist. Once AI labs no longer consume computing power on an expected scale, the price of this type of debt will drop drastically. The point is that speculators who buy these debts generally increase leverage and hold large amounts of poor quality debt.

Data compiled by Woofun AI shows that this leveraged debt structure makes the market extremely vulnerable, and any contraction on the demand side could trigger a chain reaction. The real core question is who ended up buying these debts and whether they used leverage when buying. The answer is clearly yes. Speculators use high leverage to buy these poor quality assets supported by AI narratives, laying the risk for future crises.

The ultimate successors are millions of US insurance policyholders, who are indirectly betting on the AI story and are at risk of becoming insolvent. Nick Nameth on Substack thoroughly reveals this scam: if AI-related debt is revalued at fair market value, a large portion of the US insurance industry is actually insolvent. In a global economy dominated by partial reserve banks, this brings up the central proposition of investment. Insurance companies sell life insurance and annuity policies and invest premium funds in AI-related debt to obtain high returns.

However, once AI debt depreciates, insurers' balance sheets will be hit hard, and policyholders will face huge losses. This risk transfer mechanism makes ordinary people the ultimate victims of speculative behavior by financial elites.

The US government faces a two-choice situation: it either uses national security as the ultimate purchaser of computing power, or it prints money to bail out loss-making insurance companies. Whichever path they choose, Bitcoin holders and crypto investors are winners. If the government ignores market signals and insists on investing money to develop a “silicon-based god” that is not commercially profitable, it will need to print money to fund such unproductive expenses. This will inevitably spawn more financial speculation and drive up the price of Bitcoin.

If the government chooses to bail out the insurance industry, it will print money to take on bad AI debts, expand the money supply, and then push up the price of Bitcoin. This dilemma has made monetary expansion an inevitable end, providing a solid macro foundation for crypto assets.

The political narrative revolves around confronting China and the transmission of elite interests. In the name of China, the US government can find a reason for almost any act, just like starting a global war against terrorism after the 9/11 incident. This time, the imaginary enemy being shaped is China, which provides affordable AI products. AI bosses have successfully persuaded Trump and his staff to ignore the reality that the market is proving that AI businesses don't make money, and voters' opposition to building new data centers. To defeat China, the US must implement national socialism within the capitalist system and invest more money to build a “silicon-based god.” According to this narrative, the US has the most inclusive and fair culture in the world, and must never let AGI be controlled by a country that is not a Jewish-Christian civilization. As a result, trillions of taxpayers' money were given to Elon, Sam, and Dario for xAI, OpenAI, and Anthropic R&D.

This act of arrogance is reminiscent of Icarus flying too close to the Sun, and is bound to be unsustainable.

Macro data and monetary policy show that interest rate hikes coexist with bank table expansion, and the liquidity environment is still relaxed. In June 2026, the US nominal YoY GDP growth rate was 6.6%, while the effective federal funds rate was about 3.6%. Treasury Secretary Bezent continues to issue additional short-term treasury notes. The government can earn 3% from issuing this debt, but savers will bear losses.

If the fiscal deficit is kept within 3%, the debt/GDP ratio will fall. However, the US monetary policy raised interest rates for the first time since July 2023, and the Federal Reserve voted to raise the policy interest rate by 0.25% last week. The total amount of money created by the Federal Reserve is no longer growing, and the RMP short-term treasury bond purchase program has been suspended since August 14.

If the government implements a computing power procurement plan, but the Federal Reserve does not reduce capital costs or expand the balance sheet, large-scale debt issuance will push up interest rates and arouse voters' anger. As a result, Trump and Bezent need to enlist the support of at least 7 FOMC members to ensure the plan is viable.

At the same time, commercial banks have created hundreds of billions of dollars in new currencies by expanding their total assets. What is behind this is the relaxation of liquidity regulations. After this 0.25% rate hike, banks can keep excess reserves in the Federal Reserve and receive an additional 7.5 billion US dollars in interest every year for additional loans and financial market speculation. When the two are combined, the overall effect is still stimulating, enough to support additional borrowing to invest in AI computing power construction.

Self-insurance scams reveal the operating mechanisms of private equity takeovers and false capital buffers. After the 2008 global financial crisis, the private sector was deleveraged, and the Federal Reserve lowered interest rates to close to zero. The classic game for private equity institutions is to use leverage to buy mature companies, cash out the dividends, and then go public again.

However, as the cost of capital rose, private equity bosses began searching for long-term capital pools, and insurance companies came into the market. In 2025, the asset management scale of private equity and venture capital (PE&VC) surpassed $15 trillion. Private equity bosses buy insurance companies, act as investment managers themselves, and package and sell poor quality assets to unaware policyholders. This is Captive Insurance (Captive Insurance). In order to cover up scams, regulations in some states, such as Vermont, allow original insurance companies and associated reinsurance institutions to privately set up reinsurance risk assets, and the size of the capital buffer can be set at will.

Taking Terra Luna as an example, if Do Kwon has the resources of a private equity boss, he may buy an insurance company called Alameda Insurance and use premium funds to buy USDT to stabilize the anchor. Alameda sells life insurance to Californians, holds billions of dollars, and buys investment-grade corporate bonds. Luna bought through Moody's (MCO.US) analysts, rated its corporate bonds as investment grade and gave interest that was 5% higher than the yield on 10-year US bonds.

Alameda registered Three Daggers, a reinsurance company in Vermont, and pledged only $1 of its own equity for every $100 reinsured asset. Once the USDT price fell, Luna was unable to pay interest on the bonds, the rating was downgraded, the entire structure collapsed, Alameda's book capital became insolvent, and the policyholder suffered huge losses. Most states in the US have insurance coverage limits of only $250,000 to $300,000, and the difference is unpayable, and insurance coverage funds are funded by survival insurance companies after the fact, encouraging institutions to take extreme risks.

The repetition of history indicates the inevitability of a bailout and the outlook for the crypto market. In 2008, AIG (AIG.US), as the last successor agency, absorbed a large amount of poor quality secondary CDO debt, and the government bailed out. After the TARP bailout filled the AIG (AIG.US) hole, the funds went directly to Goldman Sachs (GS.US), and Goldman Sachs (GS.US) distributed record bonuses in 2009. Paulson and Bernanke stuck to the bottom line back then and let Lehman go bankrupt, but this decision was a mistake, letting the public see how financial institutions are plundering the public.

This time around, Bezent and Walsh would never allow large insurance companies to go bankrupt and stage a financial disaster drama similar to “The Big Short.” They will continue to print money to avoid this liquidation, because after 2008, populist political power rose, and people will not be as obedient as they were in the past. Back then, Obama approved the bailout plan, and there was no large-scale prevention of home foreclosures. AOC won't be this talkative until 2028. Therefore, Walsh and Bezent must put an end to an open credit disaster.

If Trump chooses the final purchaser with poor computing power and the rating agency downgrades AI data center debt ratings, the banknotes will be printed in batches and slowly, preventing the market from fully realizing that the insurance industry is insolvent. As the founder of AI/crypto project Flop Network, Hayes believes this macro environment is very favorable. The US government will not allow the free market to stop data center construction. The original cost of computing power will drop, there will be an oversupply of spot computing power, and promote the spread of AI intelligent agents. In addition to this, a large number of new dollars will drive capital into crypto assets. In the currency expansion cycle, crypto assets perform best.