Did you hear? The big names in AI have suddenly had a change of heart, worrying about humanity’s survival—because the nearly silicon-based god they’ve been building is about to be born. In fact, it’s been close to a silicon-based god for some time now, but as the final step looms, they’ve suddenly begun reflecting on the path toward artificial general intelligence (AGI). (Note: “Silicon-God” refers to the impending super AGI in Silicon Valley slang; the author uses this term satirically, questioning whether this grand narrative is merely a fundraising and lobbying tactic.)
This is the narrative they’ve presented to the public. But being naturally skeptical, I checked the date—quarter three is about to end, and Anthropic still hasn’t gone public. I couldn’t help but wonder: what do their financials actually look like? Has the much-publicized annualized revenue growth curve already slowed? They claim the company is profitable after accounting for all operating costs. I’m eager to review their upcoming S-1 filing to understand exactly how much it costs to provide each token to users. And how many customers are actually profitable? Is this profitable customer base growing or shrinking? Unfortunately, I can’t find answers to these questions—because, they say, security comes first.
From my tone, you can tell that I believe Anthropic, OpenAI, and SpaceX are slowing down general AI development under the guise of “safety first,” not out of concern for ordinary human welfare, but due to harsh economic realities: the market is unwilling to purchase their AI products at current prices in sufficient quantities. More specifically, demand for AI is strong, but people want Chinese pricing—just one percent of U.S. prices.
When this "China-priced" challenge hit American AI professionals who saw themselves as superior, their first reaction was to cry out: "But Chinese products are of terrible quality." And when the quality of Chinese models quickly caught up, they lamented: "China merely distilled our models to create their own products." The market doesn’t care why Chinese models are cheap—it only wants the cheapest intelligent services. So these AI professionals shifted to sighing: "We care about human safety, so we’ve paused development." It sounds so noble... But this exaggerated, World Cup-style "fake injury performance" will later demand: "Yet we still need to outperform China—so the government must step in, impose regulations, and continue funding this race toward general artificial intelligence."
These three leading U.S. AI labs are using "safety first" as an excuse to slow down AGI development—a matter of critical importance to financial markets and global fiat liquidity, because these labs' demand for computing power underpins over $1 trillion in investment-grade debt and hundreds of billions in low-credit-rated loans.
Show the amount of procurement commitments made by OpenAI and Anthropic to the four major U.S. cloud providers, as a percentage of each provider’s unfilled revenue orders, to illustrate the substantial long-term orders these two AI model companies have brought to the cloud providers.
These AI labs have collectively generated no profits. Therefore, they rely on profitable tech companies like NVIDIA, Broadcom, Google, and Microsoft to provide off-balance-sheet guarantees for debts related to data center leasing and chip procurement. Subsequent purchases of chips and hardware depend on AI labs continuously training cutting-edge models—those “almost, almost, almost divine” silicon entities—while simultaneously handling inference requests for clients. But if “safety first” becomes the new guiding principle, the cost of training new models won’t vanish, but will certainly decline from current highs; companies will shift focus to improving the efficiency of power-to-intelligence conversion, meaning clients will require less computational power. Fundamentally, safety first means dismantling demand for compute.
If AI capital expenditures were funded solely by operating cash flow, there would be little to worry about. But the problem is, regardless of whether AI labs continue purchasing compute, this trillion-dollar debt still exists. Default won’t happen immediately, but once AI labs no longer consume compute at the previously anticipated scale, the value of this debt will decline. The real core question: Who bought this debt, and did they use leverage when buying it? The answer is clear—these speculators used leverage to purchase these low-quality debts. So who is ultimately left holding the bag?
Millions of American insurance policyholders have, in fact, indirectly bet on the AI story—and the “safety first” approach will harm them without any buffer. I didn’t fully understand this scheme until Nick Nameth on Substack laid it out with exceptional clarity; I’ll now explain it in simple terms for my crypto-savvy readers. The bottom line: if AI-related debt were revalued at fair market value, a significant portion of the U.S. insurance industry would actually be insolvent. This leads to the central dilemma in a global economy dominated by fractional-reserve banking: the U.S. government must choose between two options—either act as the ultimate buyer of computing power under the banner of national security, or print money to bail out deeply insolvent insurance companies.
No matter which path is chosen, we Bitcoin holders and crypto investors are winners. If governments ignore market signals and insist on pouring funds into developing this commercially unprofitable "silicon deity," they will need to print money to finance such unproductive expenditures, inevitably fueling more financial speculation and driving up Bitcoin’s price. If governments choose to bail out the insurance industry, they will print money to absorb bad AI-related debts, expanding the money supply and thereby pushing Bitcoin’s price higher.
The remainder of this article will break down this mechanism.
In the name of China
Under the guise of national security, the U.S. government can find justification for almost any action. Consider the aftermath of 9/11, when the U.S. stoked fears among its citizens and launched a global war on terror, causing immense destruction. This time, the fabricated adversary is the Chinese, who are portrayed as burying themselves in mathematics, stealing America’s cutting-edge technologies, and then selling products back to the U.S. at one percent of the price. To defeat China, one must implement state socialism within the capitalist system.
AI elites successfully convinced Trump and his advisors to ignore two realities: the market has already proven that AI businesses are not profitable, and voters across the political spectrum oppose the construction of massive new data centers and demand compensation for data theft. Since China can provide affordable AI products, the U.S. response is to pour in more funding to create that “almost, almost, almost, almost divine silicon entity.” (I’ll keep adding “almost,” because we’re just one step away from AGI—requiring nothing but belief, data theft, and taxpayer funding.)
War, economics, robotics—everything ultimately comes under the control of artificial general intelligence. Therefore, according to this narrative, other nations must use AGI according to America’s wishes. Under this logic, the U.S. is said to possess the world’s most inclusive and fair culture, and this silicon-based deity must never fall into the hands of non-Judeo-Christian civilizations, such as China. (Rolls eyes. Then rolls them again—big time.) I have my preferred place to live; others have theirs. Even if I believe my own moral culture is superior, I wouldn’t spend my entire fortune—or risk my life—to impose it on the rest of the world. You may believe American or Western culture is superior, but don’t hand over trillions of taxpayer dollars to Elon, Sam, and Dario.
The supposed superiority of the American system lies in the fact that, for the most part, hundreds of millions of informed citizens determine the prices of goods and services through free markets, with no government interference—allowing market signals to dictate what is produced and in what quantity. But now, due to national security concerns and based on the assumption that pouring vast sums of money into feeding data to predict the next token will produce AGI, the signals from the market are deemed incorrect. As a result, the U.S. government must increase its investment to develop the next generation of cutting-edge models, using cultural advantages to outpace China. This is hubris—do you remember what happened to Icarus when he flew too close to the sun?
Alright, stop talking in generalities—explain the rescue plan.
"Security first" implies a decline in computing power demand from the three major U.S. AI labs. At this point, the government could step in and enter into take-or-pay agreements to guarantee the AI labs stable profits, similar to contracts the U.S. provides to certain defense and mining companies. The government would use this computing power to advance general artificial intelligence research. Finally, the government could lease its self-developed models back to the AI labs, allowing the labs to sell inference services to clients in the U.S. and allied nations at very high prices.
This approach would allow the government to control cutting-edge models for use by the Western world as it sees fit. A problem with the private AI lab model is that these labs belong to global corporations that sometimes sell access to their models to anyone worldwide for profit, conflicting with government national security objectives. If Chinese labs rely partly on distilling American frontier models to make technological progress, and if the government directly controls R&D, China could fall behind by months or even years in the race for general artificial intelligence. Even if this scenario were plausible, it would ultimately fail—just as attempts to block the export of advanced chip manufacturing equipment failed to prevent China from developing cutting-edge chips. Information inherently seeks to flow freely. In the internet age, information control is impossible. Even before the internet existed, the United States could not prevent the Soviet Union from obtaining intelligence after successfully developing the atomic bomb. Those who believe the field of AGI will be an exception fail to understand human ingenuity and adaptability in the context of national-level strategic competition.
Funding this silicon deity requires issuing more debt. This plan is easy to sell because monetary policymakers—Treasury Secretary Bensent and Federal Reserve Chair Wash—believe AI can boost productivity. They are convinced that by fully embracing AI, the U.S. can grow its way out of its massive debt burden, a judgment that is not entirely wrong. In June 2026, U.S. nominal year-over-year GDP growth reached 6.6%, while the effective federal funds rate stood at approximately 3.6%. Bensent continues to issue more short-term Treasury bills; the government earns a 3% return on this debt, but savers bear the loss. If the fiscal deficit is kept under 3% (a significant assumption), the debt-to-GDP ratio will decline. Economic growth is primarily driven by the construction of AI data centers, fueled by the computing power demands of AI labs. From a financing perspective, as long as short-term Treasuries still yield 3%, it is financially viable for the government to act as the ultimate buyer of computing power.
There's no such thing as a free lunch. The U.S. government has run persistent deficits and relies on borrowing to fund its spending. If Wash cooperates, this matter would be straightforward. But so far, the Federal Reserve under his leadership has not been in sync with the Treasury Department led by Bessent.
Since July 2023, U.S. monetary policy has seen its first interest rate hike: at last week’s meeting, the Federal Reserve unanimously voted to raise the policy rate by 0.25%. The total amount of money created by the Fed is no longer growing; as of August 14, the RMP short-term Treasury purchase program has been halted.
If the government implements this plan but the Federal Reserve does not lower borrowing costs or expand its balance sheet, large-scale debt issuance will push interest rates higher. Rising rates on mortgages, credit cards, and auto loans will only fuel voter anger toward AI-related policies. If Trump and Bessent cannot secure support from at least seven FOMC members, the feasibility of this plan will be significantly diminished.
The above describes Federal Reserve policy from a traditional perspective, which could easily lead to a bearish view of the market. But don’t forget there’s another powerful money-printing machine: commercial banks. Walsh and Besant argue that banks should take on the responsibility of monetary expansion. Since the RMP purchases ceased on August 14, banks have created hundreds of billions of dollars in new money by expanding their total assets, enabled by relaxed liquidity regulations.
In addition to balance sheet expansion, after this 0.25% rate hike, banks' excess reserves held at the Federal Reserve will earn an additional $7.5 billion in interest annually. This money will be directed toward new lending and financial market speculation. Therefore, it’s not enough to only consider the Fed’s rate hikes and the end of balance sheet expansion—you must also account for the impact of the commercial banking system. Together, their combined effect remains stimulative. In other words, if the government chooses to do so, the liquidity environment is sufficient to support new debt issuance for investment in AI computing infrastructure.
However, if the government does not step in to purchase computing power, the debt will suffer impairment. Those who have leveraged positions in such debt will face a crisis. Below, we’ll take a closer look at this scam known as self-insurance.
Captive Insurance
I had never previously studied the insurance industry. At first, the idea that large private equity firms used assets from captive insurance companies to raise investment capital did not seem like a scam to me. But after digging deeper into how this system operated, I identified the ultimate victims.
Every credit bubble has a group of final buyers. Typically, the funds of ordinary retail investors are managed by seemingly reputable trustees. These trustees invest other people’s money and earn double returns: charging management fees while also selling their own assets to retail investors, thereby inflating the prices of their holdings. This time, the victims are policyholders who purchased U.S. life insurance and annuity products. To understand this scam, we must first examine the survival tactics devised by private equity giants after the golden age of private equity came to an end.
After the 2008 global financial crisis, the private sector deleveraged, and the Federal Reserve lowered interest rates to near zero. The classic strategy of private equity firms: identify mature companies with stable cash flows and minimal debt, apply leverage, distribute dividends to extract cash, and leave the weakened company behind in the private market. Once public market sentiment heats up, relist the distressed company to complete the cycle. In a low-interest-rate environment, this logic was entirely viable. Ordinary individuals, burdened by negative-equity mortgages and struggling to make monthly payments, lacked the capacity to increase consumption. Private equity magnates had no interest in expanding production or improving products—they merely sought to maintain existing cash flows, cut costs, and extract cash dividends for their investors.
Total U.S. credit as a percentage of GDP. Orange represents private market credit, blue represents government-related credit. After the 2008 financial crisis, private credit continued to decline while government credit steadily rose, reflecting a shift in debt structure from the private to the public sector.
The assets under management (AUM) of private equity and venture capital (PE&VC) have consistently grown since 2000, even through economic recessions, surpassing $15 trillion by 2025. The gray shading indicates economic recessions as marked by the NBER.
In the post-pandemic era, rising capital costs and the law of diminishing marginal returns have significantly pressured private equity returns. With cheap financing no longer available, firms seeking to close deals must offer higher valuations to acquire cash-flow-generating assets, causing private equity returns to decline. To secure their next fundraise, top private equity firms are turning to long-term capital pools that do not demand short-term redemptions—insurers have stepped into this role. Insurers sell life insurance and annuity policies; policyholders pay premiums, and insurers invest these funds to generate returns, paying out claims decades later. This is precisely the perpetual capital pool private equity has long sought—ideal for investing in overvalued private companies and high-yield private credit.
Thus, private equity magnates acquire insurance companies, appoint themselves as investment managers, and bundle low-quality assets to sell to unsuspecting policyholders. This is known as captive insurance.
The most absurd aspect of this scheme is how the captive insurer meets legal capital buffer requirements. Asset prices fluctuate, and regulators require insurers to maintain capital buffers to ensure policy payouts. This led to the creation of reinsurance companies, which assume the payout risk from primary insurers. Normally, primary insurers and reinsurers are independent entities, and reinsurance risk is priced at fair market value. But this framework doesn’t work for private insurance scams, whose core mechanism relies on finding a counterparty while requiring minimal自有 capital from the private entity itself. Thus, the privately held insurer established a related captive reinsurance entity—allowing the parent company to obtain reinsurance coverage with only minimal自有 capital investment.
These are regulated entities required to make regular public disclosures. Would policyholders still purchase insurance if they knew the insurer was operating such a system? To conceal the fraud, the U.S. capital system has sided with private entities. Regulations in states like Vermont contradict national prudent regulatory standards. The original insurer and its affiliated reinsurance entities can privately establish reinsurance risk assets, with capital buffers set at arbitrary levels, and once approved by state regulators, the related information is sealed and kept confidential.
There’s more detail. Before breaking down this bold scam further, let me draw an analogy from the crypto space. Do you remember Terra Luna? Luna collapsed because USDT holders sold off the stablecoin, breaking its dollar peg. What if Do Kwon had access to the resources of those New York private equity tycoons in their late sixties?
When the price of USDT fell, Luna acquired an insurance company called Alameda Insurance. Luna used premium funds to buy USDT in an attempt to stabilize the peg. Alameda sold life insurance to residents of California, holding tens of billions in assets. Alameda could not directly purchase altcoin stablecoins but could buy investment-grade corporate bonds. Luna bribed Moody’s analysts to rate its own corporate bonds as investment-grade. To attract buyers like Alameda, Luna offered an interest rate 5% higher than the yield on 10-year U.S. Treasuries—what an attractive return! Alameda registered a reinsurance company called Three Daggers in Vermont. For every $100 in reinsurance assets, Alameda pledged only $1 of its own equity. Then Alameda told regulators that the $10 billion in Luna investment-grade debt purchased with policyholder premiums was safe, and that Three Daggers would cover any losses if needed. Everything appeared fine. But as USDT continued to decline, Luna’s token price collapsed. Weeks later, Luna defaulted on its bond interest payments. Even if Moody’s analysts had accepted Rolexes to conceal the rating, they could no longer ignore the default and were forced to downgrade the debt to junk status.
If the rating is downgraded, the entire structure collapses. Regulated insurers must replenish capital after debt downgrades. But the problem is that this reinsurance asset was fraudulent from the start. Three Daggers never transferred hundreds of millions in cash to its parent company to meet capital requirements. When the scam was exposed, Alameda was insolvent on paper, leaving regulators to clean up the mess.
The policyholder will suffer significant losses. In most U.S. states, insurance coverage limits are only between $250,000 and $300,000. If your policy was supposed to pay out millions, the difference cannot be recovered. Even worse, insurance guarantee funds are financed retroactively by surviving insurers, which is entirely different from the banking insurance system. The FDIC (Federal Deposit Insurance Corporation) requires banks to pay premiums in advance. This retroactive funding mechanism indirectly encourages institutions to take extreme risks, as they do not have to pay for crises upfront.
Return to traditional finance: replace the debt of altcoin projects with private credit extended to software companies impacted by AI, and debt tied to AI data centers, where the value of such debt depends entirely on the ongoing procurement of computing power by the three major AI labs.
The column labeled “Related Reinsurance” in the table represents this fictional capital buffer, which private equity giants like Apollo, KKR, and Brookfield rely on to mask all their AI-related investments. The true quality of the captive insurer’s balance sheet assets is unknown, as they are deliberately registered in states and jurisdictions that allow concealment of actual financial data. However, based on public news reports, nearly every major debt issuance for AI data centers has these large private equity firms behind it. At the same time, they are also the largest private credit funds, investing heavily in SaaS companies. These private credit funds have already restricted investor redemptions, as these illiquid loans cannot be quickly liquidated at a discount.
Nick Nameth estimates that the scale of these fraudulent self-insurance reinsurance assets reaches $1.54 trillion. The actual amount is uncertain, but he cites the example of Brookfield’s reinsurance assets: recorded on the books at $1.48 billion, yet the entity providing the reinsurance reported to regulators that its actual liability for claims was $0.
Private credit and AI debt markets are beginning to show cracks under the weight of massive existing debt. We’ve already seen these cracks emerge before the market fully realizes the insolvency of private captive insurance companies. The trigger: downgrades to investment-grade debt securitized from AI data centers. Insurers originally purchased the highest-rated tranches, enjoying yields higher than comparable U.S. Treasuries. Once leading AI labs fail to procure computing power at expected scales, the resulting cash flows will be insufficient to service data center debt, prompting rating agencies to downgrade the securities—triggering parent companies to inject capital, while their affiliated reinsurers cannot produce cash. For investors like us who have profited from monetary expansion, the trillions of dollars in liabilities on insurers’ balance sheets will inevitably require a bailout. Baby boomers, who hold insurance policies and voting power, will vote to push for rescue measures. In 2008, AIG—as the ultimate counterparty—absorbed vast quantities of subprime CDOs, and the government bailed it out. Don’t forget: after TARP (the Troubled Asset Relief Program, the core U.S. bailout tool during the subprime crisis) filled AIG’s losses, the funds flowed directly to Goldman Sachs, which paid record bonuses in 2009. Ordinary people received foreclosure notices; the elite received generous checks.
This scene will play out again. But Bessent and Wash are not naive enough to repeat the old free-market rhetoric of the past. Back then, Paulson and Bernanke (Hank Paulson and Ben Bernanke, former Treasury Secretary and former Fed Chair) held to a firm principle: failed investments meant failure of the firm. They allowed Lehman Brothers to collapse—a decision that was wrong, and one that exposed to the public how financial institutions exploit ordinary people. This time, Bessent and Wash will never permit a major insurer to fail and reenact a financial disaster like The Big Short. They will keep printing money to avoid this reckoning. Because after 2008, populist political forces rose, and the public will no longer submit as they once did. Back then, Obama, nominally a progressive Democrat, won part of his election on rhetoric about the financial crisis and punishing bankers—yet once in office, he still approved bailout programs and did little to stop widespread foreclosures. By 2028, AOC won’t be so accommodating. So Wash and Bessent must prevent an open credit catastrophe.
If Trump selects an inappropriate final purchaser of computing power, credit rating agencies will downgrade the debt ratings of AI data centers, and money printing will be implemented gradually and slowly to prevent the market from fully realizing the insurance industry's insolvency.
History repeats itself
"Security first" won't immediately lead to massive money printing. Let Trump decide, but as a Bitcoin and crypto investor, no matter which path he chooses, more money printing will ultimately follow. This article will convince you that the market volatility following the modest rally at the end of August will soon subside. The total supply of dollars will continue to expand, driving prices of Bitcoin and select altcoins higher.
As the founder of the AI/crypto project Flop Network, this macro environment is highly favorable for me. The U.S. government will not allow the free market to halt data center construction; the raw cost of computing power will decline, leading to an oversupply of spot computing power and accelerating the adoption of AI agents. Additionally, a significant influx of new dollars will drive capital into crypto assets—during monetary expansion cycles, crypto assets perform most strongly.