There is a wave of "AI bubble theory" every six months. When is "the wolf really coming"?

Check out the Golden Kirin analyst research report for stock trading, which is authoritative, professional, timely and comprehensive to help you tap potential theme opportunities!
Just like a clock, a similar plot occurs every six months. The "AI bubble theory" always appears on time, triggering a brief panic in the market, and then quickly being overwhelmed by a new round of fanaticism.
From Goldman Sachs questioning its business returns, to China launching a highly cost-effective model, to Oracle and OpenAI launching a market-shaking US$300 billion "future contract", AI's questioning and carnival alternate.
However, Zerohedge's article analyzes that behind this cyclical debate, a deeper structural risk is emerging: the AI infrastructure race is evolving from a marathon supported by the internal cash flow of technology giants to an "arms race" that relies on external debt.
When a funding gap of US$1.5 trillion needs to be filled by the already pressured private credit market, people can't help but ask: How far is the "wolf" that will eventually come?
The first large-scale surge of concerns is now June 2024. At that time, Goldman Sachs issued a report directly pointing out whether generative AI is a capital bottomless pit that "invests too much and benefits too little", that is, a huge pit that may never bring long-term positive returns to investors. This question dropped a shock bomb in the scientific and technological community.
However, six months later, and after burning another $100 billion to "improve" the most expensive chatbot on the planet, a clear profit model still seems to have emerged in the United States.Instead, China has launched the famous DeepSeek model, which is not only open source, but also much cheaper than similar products in the United States, and requires far less expensive equipment to run than the latest Nvidia Super Graphics Card.
At the same time, there are reports that companies such as Microsoft, Google and Meta are quietly cutting back on capital expenditures. The combination of these factorstriggered the next round of selling of AI concept stocks, which began at the end of January and lasted into April.
History always seems to repeat itself. This can't help but remind people of the first Internet bubble, when those companies that were once popular eventually failed to escape bankruptcy.
'Infinite money loophole': When financing shifts from cash to debt
Time comes to September 2025, and the AI bubble is expanding at full speed and alone has pushed the stock market to its highest valuation level since the Internet bubble...
However, on September 10, Oracle shattered the tranquility of this carnival with an extremely reckless attitude.The Wall Street Journal reported at the time that it announced a five-year,$300 billion cloud computing agreement with OpenAI. This is considered one of the largest"Vendor Financing " deals in history.
Even more impactful is that Oracle reminded everyone almost simultaneously that it does not actually have enough cash to pay for a spending spree that is expected to continue into the 2030s. So, where does the money come from? Borrow.
In his latest Market Observer report, JPMorgan analyst Michael Cembalest brilliantly described this AI circular economy, which many colleagues call the "infinite money glitch."
He used a simple circular diagram to explain this phenomenon: AI companies promise to pay huge amounts of money to cloud service providers in the future → cloud service providers use this story to borrow money to build infrastructure → infrastructure is then leased to AI companies.
Cembalest pointed out that since ChatGPT was launched in November 2022, AI-related stocks have contributed 75% of the S & P 500 's returns, 80% of earnings growth, and 90% of capital expenditure growth. Data center power consumption is driving up electricity prices, for example in the PJM region, where 70% of the increase in electricity prices last year can be attributed to data center demand.
The transaction between Oracle and OpenAI is a perfect reflection of this "loophole". Doug O'Laughlin of the investment newsletter Knowledge of Manufacturing commented:
Oracle's case reveals a deeper problem: the cost of building AI infrastructure is out of control and far exceeds the technology giant's own blood-making capabilities. A report from Morgan Stanley paints this shocking picture: Total global data center-related spending is expected to reach approximately $2.9 trillion by 2028.
The report pointed out that although the internal cash flow of large technology companies is still the main source of funding, after taking into account factors such as shareholder returns,they can only raise about US$1.4 trillion at most. This means that the market will face a huge financing gap of US$1.5 trillion.
Morgan Stanley believes credit markets will play an increasingly important role in closing this gap.
Among all credit channels, private equity credit is highly expected. The bank expects that among the various types of capital to make up for the gap, private equity credit (especially asset financing) will contribute approximately US$800 billion, becoming the most important source of external funding. Consulting firm Bain later reached similar conclusions.
The future of AI seems to be deeply bound to the purse of private equity credit.
However, betting AI's future on private equity credit can be a dangerous bet. Just as the market is expecting it to "transfuse" AI, the industry's own health status has turned red.
Market data shows that the share price of BXSL, a private equity fund owned by Blackstone Group, one of the world's largest private equity credit management companies, has fallen to a new low in 2025, and its performance lags far behind the S & P 500 index. The share price of Blue Owl, another industry giant, is equally at risk. According to Bloomberg, Blue Owl has been deeply involved in financing activities in the AI field.
The plight of these private equity giants goes far beyond providing funding for data centers. They have been extensively exposed to the weakest link in the U.S. economy-consumers, especially low-income groups whose bad debt ratios (NPLs) have soared in the "buy now"(BNPL) arena.
As the Financial Review put it, the private equity industry is "sitting on a $5 trillion fear of survival." If the industry, which is seen as backing AI funding, is itself in trouble, where will the $800 billion promised to AI come from?
Bubbles within bubbles: When no one talks about bubbles anymore
While financial structural risks are becoming increasingly prominent, public discussions about the AI bubble are cooling down. Deutsche Bank analyst Adrian Cox pointed out that global online search volume for "AI bubble" has dropped 85% from its August 2025 peak. In other words, the "bubble of discussing the AI bubble" itself has burst.
But this does not mean that the alarm has been lifted. History shows that the evolution of asset bubbles is a non-linear process. In the five years before the Tech-Internet bubble burst in 2000, the Nasdaq index experienced seven corrections of more than 10%.
More importantly, in November 1998, when investment manager Michael Murphy warned that "this is a serious bubble," the Nasdaq index was less than 2000 points, and it continued to double over the next 16 months, breaking through 5000 points before finally collapsing.
With the shouts of "wolf is coming" every six months, the market seems to be tired. Oracle's huge deal reveals a dangerous signal behind the AI boom from "hard work" to "lending", while the expected financier-private credit-is itself in deep trouble.
This carnival driven by debt and dreams is even more fragile under the hard constraints of infrastructure such as power networks.
Perhaps, when everyone stopped talking about bubbles, the wolf really sneaked to the door. Or, as the old market adage goes, the market will always be irrational longer than the time you will be bankrupt.
So, are we in the midst of the biggest bubble in history? When will it rupture?
The honest answer is: No one knows. Just as Nvidia's share price hit another new high and its market value soared to a staggering $4.5 trillion, the market still bought the AI narrative.