Unlike the 2000 internet bubble, this AI cycle is entirely a heavy-asset, triangular-debt, mutually guaranteed, and massive off-balance-sheet credit affair—fully replicating the real-estate chain of the 2008 subprime mortgage crisis.
-
▶︎▶︎ The severe consequences of OpenAI and Anthropic failing to list?
OpenAI and Anthropic are expected to list on Nasdaq in October this year. In such a window for a technical bear market, if the market performance is still underwhelming at that time, it could very likely lead to their failed listing—this would be a very serious consequence.
First layer: AI giants’ cash flow dries up
OpenAI and Anthropic have posted massive losses year after year. OpenAI burns $17 billion per year and won’t be profitable until as early as 2029. If the IPO fundraising fails, they will be unable to fulfill long-term compute procurement contracts:
- Unable to pay data-center rent to cloud providers.
- Unable to fulfill their payments and Nvidia’s hundreds-of-billions GPU procurement orders.
- Private-market valuations for primary issuances collapse; subsequent financing is completely shut down—direct debt default.
Layer two: Cloud-provider debt blows up
Cloud-service providers are data-center developers. They take long-term leases from AI companies as collateral, borrow money from banks and shadow banks to build compute centers, and this is a typical pattern of lending backed by future income—exactly like subprime loans backed by future rent payments:
- AI giants default on rent; the value of the collateral goes to zero.
- Data centers under construction are left to rot; completed projects remain vacant.
- Cloud providers are unable to repay bank credit, leading to widespread loan defaults.
- Cloud providers are forced to cut capital expenditures and completely stop building new compute-power projects.
Layer Three: Nvidia gets hit twice
A revenue collapse plus massive guarantor compensation payouts—this is the core risk hub.
Operating side: Cloud providers and AI companies stop purchasing GPUs; chip orders collapse, and revenue and profits are directly cut in half. The market will immediately dump Nvidia bonds; CDS credit-default insurance prices will skyrocket; and the company’s credit rating will be downgraded.
Guarantee side: Nvidia provides debt guarantees in the hundreds of billions for AI compute-power projects. After widespread upstream and downstream defaults, guarantee-payout obligations are concentrated and settled, creating losses from off-balance-sheet contingent liabilities on the order of hundreds of billions.
Layer Four: Risk transmits to the real economy
Risk gradually transmits to the real economy across the United States through banks, shadow banks, etc.—this is the starting point of the financial crisis.
Traditional commercial banks: big players like JPMorgan and others provide project loans of the hundreds of billions to data centers. With widespread loan bad debts, bank assets shrink, credit to the entire industry tightens, and corporate borrowing costs skyrocket.
Private credit, shadow banking: This is the channel for 72% of US corporate loans. Players like Blackstone and BlackRock are heavily positioned in AI infrastructure debt. After underlying asset defaults, investors rush to redeem, forcing funds to sell all risk assets at a discount—liquidity dries up.
Pension funds, insurance institutions, and corporates: Large allocations of AI-related credit and tech stocks suffer significant write-downs; households’ retirement assets shrink, suppressing consumption. Corporate earnings decline further, leading to more debt defaults and forming a cycle of economic recession.
-
▶︎▶︎ A real-estate-style heavy-asset debt chain
More seriously, this crisis is completely different from the 2000 internet-bubble episode. In 2000 it was a pure equity-valuation bubble with light-asset dynamics and no chain-linked debts; whereas this AI industry is exactly an extreme heavy-asset model.
This round of the AI industry is completely heavy-asset, triangular debt, mutual guarantees, and massive off-balance-sheet credit—fully replicating the real-estate chain pattern of the 2008 subprime crisis.
Application companies like OpenAI and Anthropic are like house-flippers: they lock in compute power for the long term, overspend future cash flows, post huge losses every year, and keep going by raising financing.
Cloud providers like Google and Oracle are like real estate developers: they borrow money to build data centers, and take long-term lease agreements from AI companies as collateral to secure bank loans.
Chip makers like Nvidia are like building-material suppliers: they sell GPUs to cloud providers, while also providing trillion-level guarantees for AI compute-power debt. This tightly binds both sides’ credit.
Banks, private credit, pension funds, etc. are like banks that issue mortgages: they issue massive loans to data centers, and the underlying collateral is the future paid contracts from AI companies.
Currently, four cloud giants—Microsoft, Google, Amazon, and Meta—have already announced $1.1 trillion in compute-power investments; nearly $3 trillion have all been completed. Add total investment of $7–8 trillion across power and supporting industries. They are like real-estate developers, but they all rely on IPO fundraising from OpenAI and Anthropic to repay debts and keep the business alive.
-
▶︎▶︎ Why can OpenAI and Anthropic bring down the entire debt chain?
OpenAI’s forward compute-power contract of $750 billion and Anthropic’s procurement orders in the hundreds of billions are the core pillars of cloud providers’ and Nvidia’s future revenue.
The two companies have massive, persistent losses and no stable operating cash flow. Without a route to equity financing through public companies, they can only keep going by raising money in the primary market. All long-term compute contracts and infrastructure loans bet on fulfilling payments via large IPO fundraises.
Cloud services companies borrow money from banks to build data centers, with collateral being many years of rent from OpenAI and Anthropic. If the IPO fails or fundraising falls short, the AI companies will be unable to pay rent; the data centers sit vacant, and the cloud providers can’t repay their bank loans—resulting in trillion-level credit losses.
Nvidia provides large debt guarantees for compute-power projects while also holding massive GPU order books. When AI companies default and cancel procurement, Nvidia’s revenue crashes, yet it still has to pay out on its guarantees—its own credit blows up.
Orders for power, equipment, and data-center supporting businesses shrink. Large numbers of projects are halted and left unfinished. All $0.7–0.8 trillion in forward investments can’t be realized, leading to a credit collapse across the whole industry.
In summary, OpenAI and Anthropic are the two largest paid endpoints in the trillion-dollar debt chain, and they are also the only two that have clearly and independently listed on Nasdaq, making them the market’s only centralized window for raising large amounts of capital.
As long as the two companies’ IPO fundraising falls short of expectations and the funding chain breaks, the front-end cloud giants’ $1.1 trillion already invested in data centers, the $3 trillion projects awaiting construction, and investment across the entire industry supply chain will all see debt defaults—which will trigger a financial tsunami like the subprime crisis.
-
▶︎▶︎ How do we seize this opportunity?
If a financial crisis really breaks out, of course I hope it won’t. But if it really does break out, where is our opportunity?
Risk transmission order: Listed valuations miss expectations → primary-market AI unicorn valuations collapse → cloud providers’ compute-power long-term contract valuations decline, and banks’ data-center loan defaults rise → Nvidia’s orders decline plus pressure to pay under guarantees → the Nasdaq AI “group hug” breaks apart, compounded by a negative gamma selloff with no support, leading to a sharp drop → the dollar weakens; safe-haven demand rises for gold and the Japanese yen.
Profit can be divided into four main directions: short the AI compute-power chain, hedge downside via index puts, go long on safe-haven assets, and bet on a decline in OpenAI and Anthropic valuations via pre-IPO positions.
Direction One: Short the entire AI compute power debt chain
This is the main source of profit when the rally explodes. The core short thesis is 【Nvidia $NVDA】. It is both the担保方 across the entire chain and the exclusive supplier of chips. After AI capital expenditures contract, orders crash. Off-balance-sheet guarantee bad debts will hit profits twice. The large bear Burry has been continuously adding to Nvidia put options.
Low risk: You can buy forward put options, covering the listed dates of OpenAI and Anthropic.
High risk: You can short via contracts, but the risk of losses is even higher. It’s suitable to enter only when certainty is higher at that time.
Other major short targets:
Compute-power cloud providers: Microsoft $MSFT, Amazon $AMZN, Google $GOOGL, Oracle $ORCL
. All $1.1 trillion of compute-power infrastructure depends entirely on rental cash flows from OpenAI and Anthropic.
Semiconductor compute-power ETFs, second-tier chips: Invesco Philadelphia Semiconductor ETF $SOXX, Micron $MU, SanDisk $SNDK
. Storage demand is completely tied to AI data center construction; when capital expenditures contract, inventories pile up and prices crash.
Small and mid-cap AI US stocks: C3. ai $AI, SoundHound $SOUN
Etc. Without real cash flow, this is pure AI-concept speculation. Liquidity is poor; when the bubble bursts, it can retrace 80–90%.
You can also short other related US stock indexes or the US dollar.
Direction Two: Go long on safe-haven assets
1) Gold (the most robust)
With Fed policy swinging and rising credit risk in US equities, the dollar is likely to weaken in the long run. Gold’s core level keeps moving higher, and when the AI bubble bursts, it will trigger global safe-haven buying demand.
2) Japanese yen
On 7.30, the Japanese yen’s single-day surge of 2.5% has already validated its safe-haven attribute. With systemic risk in US equities, global funds sell off dollar assets to convert into yen safe-havens. Combined with interventions by the Bank of Japan to support the exchange rate, you can directly go long the Japanese yen.
Direction Three: Trade based on market expectations
You can place an early bet on OpenAI or Anthropic’s IPO valuation collapse, if you can predict suitable underlying targets in the market.
-
▶︎▶︎ Conclusion
Since July 13, I have already issued signals more than five times that a variety of potential financial-crisis scenarios could occur. I urge investors to take them seriously—don’t dismiss it as baseless rumors or attention-seeking hype.
Cherish the whistleblowers who run against the tide. Markets rise and fall—don’t blindly follow; be in awe. Even if you don’t plan to short, try to do as little investing as possible in this window. It’s not too late to act once the market warms up or the crisis is resolved.
The above are my personal thoughts and do not constitute any investment advice.

