The original text is from Ed Zitron
Translation｜Odaily Planet Daily Qin Xiaofeng（@QinXiaofeng888）
Editor's note: OpenAI secretly submitted its IPO application to the SEC in June 2026 and is gradually approaching its IPO, but the current timeline may be pushed back to 2027; if successful, OpenAI will become one of the largest IPOs in the AI field.
Recently, independent technology critic Ed Zitron published a lengthy article criticizing OpenAI. He pointed out that the real AI bubble is essentially the "OpenAI bubble" (OpenAI is the bubble); if OpenAI fails, it will become the "Lehman Brothers" of the AI era, potentially triggering a repricing of data centers, AI infrastructure, and tech stocks, emphasizing that OpenAI's business model is unsustainable, reliant on subsidies and cyclical financing, lacking genuine demand. (Odaily note: Ed Zitron is known for his blunt and financially detail-focused style in dissecting tech bubbles. This long article is the latest radical piece in Zitron's long-term critique of the AI bubble, following several exclusive reports such as OpenAI's losses, the Silicon Valley Bubble series, and more.)
The original text is long, and Odaily Planet Daily translates as follows, Enjoy～
——————————————
Today's article is one of the largest free newsletters I have ever written, gathering the results of the past six months of work.
And it all starts with a question: How much do you trust Sam Altman? The stock market, and to a certain extent the global economy, depends on your answer.
OpenAI has become one of the largest liabilities in modern economic history. You could argue that OpenAI is no longer the center of the AI bubble—you can talk about open-source models or Anthropic or any other element—but without OpenAI, there would be no AI industry, and the rationale for tens of trillions of dollars in capital spending would be utterly obliterated.
The AI bubble does not stem from any actual investment returns—whether in purely monetary terms like revenue or profitability, productivity gains, or any tangible or measurable benefits. Instead, it is a kind of cult-like psychosis that has infected the minds of some of the most powerful and wealthy individuals and institutions, a strong mythos around one company that has inspired— and been used to inspire—the largest capital misallocation in history.
While this may annoy some people, I firmly believe that the only reason this has lasted so long is that OpenAI has not yet collapsed. Its failure will be a watershed moment—the Lehman Brothers of the AI bubble, an event that will define the end of one era and the beginning of another, waking those who are infected from their delusion. Without this wake-up call, NVIDIA continues to sell GPUs, the semiconductor industry's coffers continue to swell, and more and more spending commitments are being made.
OpenAI plans to burn through more than $852 billion by the end of 2030. It constitutes $748 billion of the remaining obligations of Microsoft, Amazon, and Oracle, plus at least $70 billion in RPOs (Remaining Performance Obligations) from Cerebras, CoreWeave, Nebius, IREN, Lambda, and Nscale (according to Kakashii’s data), and plans to spend an uncertain tens of billions more on Broadcom’s “Jalapeño” chips. It plans to spend $50 billion or more this year on compute power, which I estimate exceeds 50% of the total global AI spending (OpenAI accounts for more than 50% of total AI compute infrastructure).
The only reason OpenAI can afford these expenses is thanks to its latest round of financing (assuming it is fully completed) totaling $122 billion, with at least $50 billion already secured, including $20 billion from SoftBank (a total of $30 billion, with a third tranche maturing on October 1, 2026). NVIDIA mentioned in its latest quarterly earnings report that it “estimates that an AI research and deployment company contributed a significant share to [its] revenue for the first quarter of fiscal 2027 by purchasing cloud services from it,” referring presumably to OpenAI.
The AI bubble is the OpenAI bubble—heading towards an inevitable end
OpenAI is the reason everyone is paying attention to AI. In March 2019 (according to JustDario’s data), NVIDIA acquired a company called Mellanox, which produced the high-speed networking technology necessary for creating AI GPU clusters. Four months later, Microsoft invested $1 billion in OpenAI and began purchasing AI GPUs and building AI infrastructure for it. By March 2020, NVIDIA shipped its A100 GPUs, and in May 2020, Microsoft announced it had built a supercomputer for OpenAI with “over 285,000 CPU cores [and] 10,000 GPUs.”
ChatGPT was launched in November 2022, arriving at just the right time for the tech industry—by then, the tech industry was running out of ideas and was teetering on the brink of a long-term recession. The IPO market had collapsed, interest rate hikes ended the era of zero rates, and the overhiring during the pandemic began to culminate in some of the most severe layoffs in industry history, while global venture capital shrank after a historic over-investment in 2021, leaving tech stocks battered.
For the first time, the tech industry was forced to live within its means—something it had never been willing to do. Large tech companies were out of favor with investors and the public. The excesses of the past decade—combined with the growing frustration towards the “tech exceptionism” (the belief that the rules governing other parts of the world didn't apply to Silicon Valley)—have tested the patience of regulators and legislators. Moreover, without “one more thing”—a sensational, game-changing product category—it no longer had the excuse for extravagant spending or regularly breaking the implicit and explicit rules governing society.
OpenAI’s existence lent legitimacy to a frenzied and lavish era. The behemoths were starved of new hyper-growth ideas, so they pointed to ChatGPT having “the fastest-growing user base of all time” and the Microsoft “supercomputer” that built it, telling investors they would be left behind if they didn’t invest. Amazon, Meta, and Google announced their own vague “supercomputers” in 2023.
By the end of 2023, NVIDIA had sold half a million A100 GPUs, and the only reason it was able to do this was due to the rapid growth of ChatGPT. Sam Altman’s brief departure would only further inflate the AI bubble—adding a layer of dreary court intrigue to an industry that lacks innovation and personality—and further solidified Microsoft’s role as the parental benefactor of OpenAI, ensuring Altman’s return to the helm.
To clarify, when I say “rapid growth,” I mean OpenAI reached 100 million weekly active users by the end of 2023, generating monthly revenues of around $108 million. Microsoft invested another $10 billion that year, with most of the funding being provided in the form of Azure credits.
OpenAI is also the reason Anthropic exists—not just because several of its founders came from that company, but also because both Google and Amazon agreed to provide it with a total of $6 billion in 2023, as a means of competition “with” Microsoft’s new darling, giving both ample reason to spend billions more “to ensure they don’t miss AI.”
When you take the word “AI” out of the equation, all of this seems a bit absurd. $16 billion in equity investments, coupled with over $150 billion in capital expenditures by the end of 2023, are all essentially rationalized because of the popularity of one website.
The only reason these two companies have been able to grow is that the behemoths have financed their entire infrastructure.
In the fourth quarter of 2023, global venture capital dropped to its lowest level since the third quarter of 2016, with American startups taking home $183.6 billion of the total investment that year. Venture capital itself simply cannot—and will not—really support them at the scale of the infrastructure required by OpenAI or Anthropic, nor would there be any appetite for financing coming from big tech or the debt financiers of data centers if the value of these two companies weren’t inflated nearly entirely due to OpenAI’s success.
If you were to take OpenAI out of the picture from 2020 to 2024, the AI bubble would not have inflated at all. No other major AI company showed any signs of life—be it those propped up by the behemoths, funded by venture capitalists, or those launched by other tech firms.
The only reason any super large company’s AI efforts generate revenue—besides OpenAI and Anthropic, the revenue is pretty meager!—is that they know all they have to do is sit there and keep saying “AI is the future” until customers eventually capitulate and try it... largely because everyone else is talking about ChatGPT.
Anthropic was only regarded as a runner-up until early 2025, and its ongoing financing was only possible because people wanted to invest in the next OpenAI; its initial rounds of financing and infrastructure were justified solely in its attempt to compete with OpenAI.
Will U.S. data center debt trades hit $178.5 billion by 2025? Almost entirely vouched for by OpenAI’s growth and its greedy demand for compute power, as aside from OpenAI (and later Anthropic), no one else was using massive clusters of tens of thousands of GPUs, and it didn’t seem that this scale of compute market was going to emerge in the months and even years following.
The biggest consumers of compute power are still Microsoft (for OpenAI), Google (for Anthropic), Amazon (for OpenAI and Anthropic), CoreWeave (for OpenAI and Anthropic), Meta (following the lead of other behemoths), and Oracle (for OpenAI). Beyond that, there is almost no evidence—and I did look carefully—that AI compute demand exceeds a few billion dollars, which is already generous.
All of these investments—whether in AI startups or data centers—exist to fund the next OpenAI, or to become the landlord of the next OpenAI.
Assuming—because no one has ever seriously thought about it—that since there is one OpenAI, there will be more OpenAIs blossoming. Since there is one large compute customer existing, the template for the next compute-heavy startups has been set... and, once again, because no one has ever really thought about anything, no one realizes that the reason there hasn’t been a second OpenAI is that OpenAI and Anthropic are financially psychological warfare meticulously crafted by the world’s largest software companies.
OpenAI and Anthropic are super scale psychological warfare experts built solely for the Silicon Valley single culture
The brutal reality is that you cannot fund an AI lab through venture capital. While OpenAI and Anthropic have raised nearly $300 billion over the past few years, their actual infrastructure costs—powering their services with GPUs and data centers—have been entirely funded by the behemoths, likely racking up another $250 billion in the process, as Microsoft has indicated that by early 2026 it had spent $100 billion on its relationship with OpenAI.
However, the real cost is not just financial; it also includes the experience and industry knowledge required to execute massive infrastructure bailouts. No company other than Google, Microsoft, and Amazon has the scale or build experience of the kind of AI cluster that OpenAI (and later Anthropic) needs.
We know this for several reasons. First, because prior to 2023, hardly any company was building AI compute clusters at the scale required by OpenAI or Anthropic. The closest might be cryptocurrency mining companies, and it is telling that today many new cloud service providers (most notably CoreWeave) were originally operating warehouses filled with ASIC chips to mine Bitcoin and Ethereum.
Second, because, according to discussions with industry insiders, the entire Overton Window regarding what constitutes a “large” facility has shifted. Previously, a 50MW data center would have been considered a significant (even noteworthy) development project. These are exceptions rather than the norm, with most data centers being much smaller. The only companies with experience building at that scale are mostly the behemoths.
By portraying OpenAI as a "venture capital-backed startup", the behemoths have created an illusion that this is the next category of big company, which will in turn create the next massive center of demand for cloud computing—except the only reason these companies exist is that the behemoths themselves are facilitating their existence with immense funding, allowing them to burn cash without ceasing.
This is why the idea that OpenAI will continue to grow infinitely is the core of the AI bubble myth. One OpenAI allows others—no matter how illogical—to imagine more OpenAIs coming to existence, which in turn means those OpenAIs will need just as much compute power as OpenAI.
Foolish investors who believe this nonsense can rationalize it through countless buy-side analysts or complicit media figures—they talk about the “insatiable demand for compute”, point to capacity constraints (wrought by slow data center buildouts, and the fact that OpenAI and Anthropic are consuming most of the global compute) and GPU price increases as evidence of strong underlying demand, while never digging into it.
The biggest trick the behemoths play is to never give an inch. By committing over a trillion dollars toward AI capital expenditures with no dollar in profit shown, they provide justification for anyone investing in AI data centers, with the underlying logic that “the largest companies in the world can’t all be wrong”—even if the reason they are doing so is to expand capacity for OpenAI and Anthropic, two companies the behemoths themselves are hatching.
To spend so much on AI infrastructure by the behemoths is fundamentally illogical and insane, and the reason so few people are willing to say so is that until recently, to imply that this was a waste of money was also seen as radical—almost entirely due to OpenAI’s existence and ongoing growth. (Note: I recognize that Anthropic has gained significant attention and grown rapidly this past year, but it has done so due to A) the myth of OpenAI, and B) because it is also being hatched and allowed to operate at large losses.)
Whether or not you get any utility from LLMs is irrelevant, as that’s not the actual foundation for data center investments. While the accelerated progress in code generation (which itself could only happen with gigantic subsidies) may have helped Anthropic grow, the vast majority of data center capital expenditures are chasing the illusion of what AI might become, rather than having any connection to the overall revenue of the companies or their economic condition—of course, apart from their compute expenditure.
This is the underlying greed driving this wasteful, reckless, and ruinous era—believing that there will be another OpenAI, as well as, as I said, the opportunity to be the landlord of the next OpenAI. And because the media and analysts seldom have original thoughts, everyone contextualizes (and continues to contextualize) this waste through the same tired clichés, saying it’s “just like Uber (it’s not!)” or “just like Amazon Web Services (who spent $29.7 billion on capital expenditures between 2003 and 2015, adjusted for inflation).”
免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。
