「雷曼时刻」在逼近？大空头1.5万字长文起底：OpenAI必将崩溃，全球股市恐遭清算

Original Author: Li Dan

Original Source: Wall Street News

As OpenAI moves closer to its IPO, a blog post of approximately 15,000 characters has once again brought the debate over the AI bubble to a fever pitch.

Ed Zitron, a commentator who has long been bearish on AI and commands a large readership in the tech industry, has put forward his most radical proposition yet in a recent blog post: The real AI bubble is, in essence, the "OpenAI bubble." If OpenAI ultimately fails, it could become the "Lehman Brothers" of the AI era, not only shattering the entire investment logic of AI but potentially triggering a massive repricing of data centers, AI infrastructure, and even global tech stocks.

His views quickly caught the attention of financial media. In their view, Zitron's core argument isn't whether AI has value, but whether OpenAI possesses a business model capable of sustaining the entire AI capital cycle. If the answer is no, then the financing, computing power investments, and capital expenditure systems built around OpenAI could face a chain reaction.

Of course, this is not the market consensus. Investors like Howard Marks, co-founder of Oaktree Capital, have recently stated that, compared to earlier suspicions that AI might just be a bubble, they now increasingly recognize AI's long-term value as a General Purpose Technology, believing the industry is still in its early commercialization stage.

AI Bubble, or OpenAI Bubble?

Unlike most "AI bubble theories," Zitron makes a more impactful judgment:

What truly warrants concern is not the entire AI industry, but a single company.

In his view, since the launch of ChatGPT at the end of 2022, OpenAI has effectively become the "credit anchor" of the entire generative AI era.

Investors are willing to believe that: AI will change the world; hyperscale data centers are worth building; GPU demand will grow rapidly for the long term; giant model companies will eventually become profitable; AI startups can create sufficient end-user demand.

All of this, according to Zitron, is predicated on OpenAI's continued rapid growth. He argues that OpenAI has not only defined the current AI boom but has also shaped the capital market's valuation logic for the entire AI industry chain. Therefore, if this core assumption is broken, the impact could far exceed that of a single unicorn company.

In other words, OpenAI is no longer just a company; it has become a "systemically important institution" for the entire AI investment cycle.

Why Does He Believe OpenAI's Business Model Has Fundamental Flaws?

Zitron's critique primarily focuses on three areas.

First, inference costs remain too high.

As ChatGPT's user base continues to grow, every user query incurs ongoing costs for GPUs, electricity, and servers. If a large number of users remain permanently on low-cost or even free tiers, and enterprise revenue growth cannot simultaneously cover these costs, then scaling up might actually mean expanding losses.

Second, capital expenditure far outpaces cash flow improvement.

Currently, the biggest expenses in the AI industry are no longer model training, but inference computing, GPU procurement, and global data center construction.

OpenAI and its partners are driving tens of billions, or even larger, of dollars in data center investments, which typically take years to recoup. If future AI demand growth falls short of expectations, a significant amount of infrastructure could face underutilization issues.

Third, continued reliance on external financing.

Zitron analyzes that he believes OpenAI will require ongoing financing for years to cover expenses related to model R&D, computing power procurement, and infrastructure construction. If capital market risk appetite declines or the financing environment tightens, its business model will face greater pressure.

These views remain Zitron's personal judgment and have not been endorsed by OpenAI, but they do reflect the recent market debate surrounding the return on investment (ROI) for AI.

Why Oracle, CoreWeave, and Data Center Operators Are in the Spotlight

More than OpenAI itself, Zitron is concerned about the leverage effect within the industry chain.

Over the past two years, the U.S. tech industry has witnessed an unprecedented wave of data center construction.

Hyperscalers like Microsoft, Google, Meta, and Amazon have all ramped up their capital expenditures. Meanwhile, companies like Oracle and CoreWeave have taken on an increasing share of AI computing infrastructure projects.

These projects heavily rely on: long-term leases, project financing, private credit, corporate bonds, and massive capital expenditures.

If future demand from core clients like OpenAI falls short of expectations, or if capital markets reassess AI returns, data center asset utilization, lease contracts, and even financing capabilities could be affected.

Media reports indicate that Zitron believes if OpenAI encounters a major setback, companies like Oracle and CoreWeave that depend on growing AI infrastructure demand could be the first impacted. The high valuations the market previously granted these companies were largely predicated on the expectation of sustained explosive AI demand.

Of course, tech giants like Microsoft, Meta, and Alphabet continue to expand their AI capital expenditures and generally emphasize that AI infrastructure investments align with their long-term strategies. Therefore, there are currently no signs of a widespread pullback in capital spending.

Why Are Anthropic and SoftBank Also Drawn into the Discussion?

Beyond OpenAI, Zitron also targets Anthropic.

His reasoning is that, although the two companies pursue different development paths, they share common characteristics: a constant need for massive capital investment to build models and procure computing power, and reliance on large tech companies for computational resources and financial support. If the pace of AI commercialization slows, both companies could face profitability pressures.

Another entity frequently mentioned is SoftBank.

In recent years, SoftBank has returned to the forefront of major AI investments, actively participating in financing for AI infrastructure, chip, and model companies.

If the AI industry enters a valuation adjustment cycle, SoftBank's vast AI asset portfolio would naturally become a focus of market attention. However, for now, SoftBank remains firmly committed to AI's long-term development, viewing it as a key direction for the next technological revolution.

Has the AI Trade Already Become Overheated?

In fact, Wall Street has been debating whether AI is entering a bubble phase for over a year.

Those supporting the "bubble theory" argue:

• Investment in AI infrastructure is growing much faster than revenue;
• The profitability model for large language models has not been fully validated;
• Data center capital expenditure is at an all-time high;
• Market valuations increasingly rely on growth expectations for years to come.

The optimists, however, believe that AI is a classic General Purpose Technology revolution, similar to the internet and electrification. Initial investments often far exceed short-term returns, but they create entirely new industries and business models in the long run.

Howard Marks recently stated that he has shifted from initially suspecting AI might just be a bubble to increasingly acknowledging its long-term value. He argues that modern AI's demonstrated reasoning, contextual understanding, and interactive capabilities are unprecedented, and therefore cannot be simply analogized to historical speculative bubbles.

Some academic studies also offer more neutral conclusions: the current AI market features both genuine technological progress and localized valuation overheating with premature capital expenditure. Thus, it is more akin to a "technological revolution overlapped with localized bubbles" rather than pure speculative frenzy.

What's Truly Worth Watching Isn't Whether OpenAI Will Fall

Regardless of whether one agrees with Zitron's judgment, the questions he raises are becoming a focal point for a growing number of investors:

When will AI spending actually translate into stable cash flow?

Over the past year, the capital market almost defaulted to the idea that higher AI capital expenditure was always better.

But recently, whether it's chip stocks, server manufacturers, or cloud computing companies, investors are paying closer attention to another set of metrics: enterprise AI revenue growth; AI product paying rates; the rate of decline in inference costs; data center utilization rates; and the ROI cycle of AI investments.

If these metrics continue to improve, the current massive capital expenditures may ultimately prove to be a forward-looking investment akin to the internet era. However, if the pace of commercialization lags behind investment expansion persistently, the market's valuation logic for the AI trade may also face recalibration.

Therefore, what Ed Zitron's lengthy article truly sparks discussion about is not "whether OpenAI will definitely become the next Lehman Brothers," but that it once again places the most critical question of the AI era before investors: after capital expenditure continues to set new records, can cash flow and profitability keep pace? The answer to this question will likely determine the true direction of the global AI trade in the coming years.