Every six months, there’s a wave of “AI bubble talk”—when will the “wolf really come”?

```

Just like a clock, every six months, a similar plot unfolds again. The "AI bubble theory" always appears on time, causing a brief panic in the market, only to be quickly drowned out by another wave of mania.

From Goldman Sachs questioning its commercial returns, to China launching highly cost-effective models, and then Oracle and OpenAI dropping a market-shaking $300 billion “future contract”, skepticism and euphoria about AI alternate on stage.

However, as Zerohedge analyzes, behind this cyclical debate, a deeper structural risk is emerging: the race for AI infrastructure is evolving from a marathon supported by tech giants’ internal cash flows into an “arms race” dependent on external debt.

When a funding gap as high as $1.5 trillion must be filled by an already pressured private credit market, one can’t help but ask: how far off is that “wolf” that will eventually arrive?

"The AI Bubble Theory" Plays Out on Time

The first large wave of concern hit in June 2024. At that time, Goldman Sachs issued a report directly pointing to generative AI as possibly being a bottomless pit of capital—“too much investment, too little return”—a pit that may never bring long-term positive returns to investors. This skepticism sent a shockwave through the tech world.

Yet, six months on and another $100 billion spent “improving” the world’s most expensive chatbot, a clear profit model still seems absent in the US. Instead, China launched the much-hyped DeepSeek LLM, which is not only open-source but also cheaper than its US counterparts and requires much less expensive hardware than the latest Nvidia supercards.

Meanwhile, reports surfaced of Microsoft, Google, and Meta quietly scaling back capital expenditures. These factors combined to spur the next wave of sell-offs in AI concept stocks—a sell-off lasting from late January to April.

History seems to repeat itself. This inevitably brings to mind the early days of the Internet bubble, when once-popular companies ultimately couldn’t escape bankruptcy.

"Infinite Money Glitch": When Funding Moves from Cash to Debt

The timeline reaches September 2025. The AI bubble is at full steam, single-handedly driving the stock market to its highest valuations since the Internet bubble...

However, on September 10, Oracle shattered the carnival’s calm in an extremely bold move. According to the Wall Street Journal, it announced a five-year $300 billion cloud computing deal with OpenAI. This is seen as one of the largest vendor financing deals in history.

More strikingly, Oracle almost simultaneously reminded everyone—it actually didn’t have enough cash of its own to pay for this spree expected to last into the 2030s. So, where does the money come from? Borrowing.

JPMorgan analyst Michael Cembalest, in his latest “Market Watch” report, incisively described this AI circular economy, which many peers call the “infinite money glitch.”

He explained the phenomenon with a simple cyclic diagram: AI companies promise huge future payments to cloud providers → cloud providers use this story to borrow for infrastructure construction → infrastructure is then leased back to AI companies.

Cembalest points out that since the launch of ChatGPT in November 2022, AI-related stocks have contributed 75% of S&P 500 returns, 80% of earnings growth, and 90% of capex growth. Data center power usage is pushing up electricity prices—for example, in the PJM region, 70% of the price increase last year was due to data center demand.

And the Oracle-OpenAI deal is a perfect embodiment of this "glitch." Doug O'Laughlin of investing newsletter Fabricated Knowledge commented:

Oracle simply can't pay for all this with cash flow—they must issue shares or go into debt to realize their ambitions... a stable oligopoly is breaking apart... what was a disciplined race funded by cash flow may now become a debt-driven arms race.

$1.5 Trillion Financing Gap—Can Private Credit Fill It?

Oracle’s case reveals a deeper issue: the cost of building AI infrastructure is running out of control, far surpassing the self-productive ability of tech giants. A Morgan Stanley report paints this shocking picture: by 2028, global data center-related spending will reach about $2.9 trillion.

Although internal cash flow of big tech companies is still the main source of funds, after factoring in shareholder returns, they can self-raise at most about $1.4 trillion. This means there will be a massive $1.5 trillion funding gap.

Morgan Stanley believes that to close this gap, the credit market will play an increasingly important role.

Among credit channels, private credit carries high hopes. The bank expects that among all types of capital to plug the gap, private credit (especially asset financing) will contribute about $800 billion, becoming the most important external funding source. Consulting firm Bain later reached similar conclusions.

The future of AI now seems deeply bound to the money bags of private credit.

Private Credit: AI’s “Savior” or “Achilles Heel”?

However, betting the future of AI on private credit could be a dangerous wager. Just as the market looks to it for “blood transfusion” into AI, the industry’s own health is flashing red.

Market data shows that BXSL, one of Blackstone’s private credit funds and among the world's largest, has fallen to a new low in 2025, far underperforming the S&P 500. Another industry giant, Blue Owl, also faces danger. According to Bloomberg, Blue Owl has already become deeply involved in AI funding activities.

The issues confronting these private credit giants go well beyond simply funding data centers. They’re deeply exposed to the weakest links of the US economy—consumers, especially low-income groups where “Buy Now, Pay Later” (BNPL) non-performing loan rates are soaring.

As the Financial Review put it, the private equity industry is “sitting on $5 trillion of existential dread.” If this sector—regarded as the AI funding backstop—falls into trouble, then where will the promised $800 billion for AI come from?

The Bubble Within the Bubble: When No One Talks About Bubbles

As financial structural risks grow, public discussion about the AI bubble is cooling off. Deutsche Bank analyst Adrian Cox notes that worldwide internet searches for "AI bubble" have dropped 85% from a peak in August 2025. In other words, “even the bubble of the AI bubble discussion” has burst.

But that doesn’t mean the alarm is off. History shows that asset bubbles do not deflate linearly. In the five years prior to the 2000 dot-com crash, the Nasdaq experienced seven pullbacks of over 10%.

More importantly, in November 1998, when fund manager Michael Murphy warned that "this is a serious bubble," the Nasdaq was below 2000 points, but over the next 16 months, it doubled and after breaking 5000, then finally crashed.

When Will the Wolf Come?

With each six-month "wolf is coming!" call, the market seems to be getting fatigued. Oracle's huge deal exposes the dangerous shift in the AI boom from “doing the work” to “borrowing money,” while the expected backer—private credit—is itself mired in trouble.

This revelry driven by debt and dreams is made even more fragile by the hard constraints of infrastructure such as power grids.

Maybe, when no one talks about bubbles anymore, the wolf will really have crept to the door. Or, as the old market adage goes: The market can remain irrational longer than you can remain solvent.

So, are we now in the greatest bubble in history? When will it burst?

The honest answer: nobody knows. As Nvidia’s stock price hits new highs and its market cap soars to an astonishing $4.5 trillion, the market is still buying into the AI narrative.

Risk Warning and DisclaimerThe market carries risks. Investment requires caution. This article does not constitute personal investment advice, nor does it take into account individual users’ unique investment goals, financial situations, or needs. Users should consider whether any opinions, viewpoints, or conclusions in this article are suitable for their specific circumstances. Investment decisions made based on this are solely at your own risk. ```