Anthropic secretly submitted an IPO application, closely followed by OpenAI, as the two AI giants compete to open the public market.

Written by: Zhao Ying

Source: Wall Street Observer

Recently, Anthropic secretly submitted an IPO application, with OpenAI following closely behind. The two AI giants are vying to open the public market. However, behind the impressive revenue figures, a more concerning question is emerging: Is this IPO boom a genuine extension of technological benefits to the public, or is it a carefully planned exit for early investors to transfer risks to ordinary people?

According to analyst Alberto Romero from The Algorithmic Bridge, Anthropic’s annualized revenue run rate is close to $50 billion, and it is expected to achieve its first profitable quarter; OpenAI’s revenue scale is slightly lower, with no timeline for profitability — both companies are in a constant cash-burning state. Given their current scale, continuing losses are no longer a sustainable business model.

Michael Hartnett, chief strategist at Bank of America, holds a sharp judgment: this IPO wave is less about investment opportunities in the public market and more about early investors transferring years of accumulated risks to the public market as an exit strategy. Ray Dalio, founder of Bridgewater, once said, “The process of bursting a bubble is the process of converting wealth into cash”— for insiders, whether a bubble exists or not, they are the winners.

At the same time, the commercial reliability of AI technology is facing increasing skepticism. Microsoft announced this month that it is canceling the Claude Code authorization and switching to internal tools due to unsustainable costs; Uber has set a cap of $1,500 a month on employees’ consumption of AI programming tool tokens; Starbucks announced a withdrawal from AI tools in North America just nine months after implementation, citing reliability issues. Even OpenAI CEO Sam Altman acknowledged that AI costs have become “a huge problem.”

The IPO window suddenly narrows, giants race to be first

Recently, Anthropic announced in a blog post that it has secretly submitted an S-1 draft to the U.S. Securities and Exchange Commission, and once approved, the company will have the option to formally launch its IPO. OpenAI's IPO process is also accelerating. Alberto Romero pointed out in his analysis that if Musk's SpaceX acquisition of xAI is included, the market is actually facing a “triple IPO” impact.

This timing is intriguing. Both companies are currently operating at a loss, yet their valuations are astonishingly high, with large capital expenditures and uncertain returns, while also carrying massive long-term spending commitments. Alberto Romero believes the essence of this race is not about technological competition between the two companies, but a capital game that is running against time— the IPO window is narrowing not because AGI (Artificial General Intelligence) is imminent, but because the shelf life of the narrative “AGI is coming” is rapidly expiring.

Risk transfer logic: private gains, socialized losses

Alberto Romero cites Michael Hartnett's view that the core mechanism of this IPO lies in the systematic transfer of risk. Early investors use the IPO to sell years of accumulated high-risk exposure to pension funds, index funds, and a broader public of investors.

Using Ray Dalio’s framework to understand: whether or not a bubble exists, the outcome for insiders is largely the same— if there is no bubble, they continue to operate as usual and eventually become wealthy; if there is a bubble, they exit early and cash out immediately. The ordinary investors who take on the tail risks are those who buy in at the secondary market.

Alberto Romero summarizes this logic as “privatized gains, socialized losses.” He points out that if the market realizes that the current valuations are inflated before the IPO, the opportunities for Anthropic and OpenAI to cash out at a high will be significantly discounted; however, if the market makes corrections only after the IPO is completed, the losses will spread throughout the economy and be shared by ordinary people.

Corporate trust shaken, AI commercialization encounters real challenges

The above concerns are supported by a series of real signals from corporate clients. Alberto Romero cites multiple recent cases in his article: Microsoft canceled the Claude Code authorization and shifted to internal tools due to unsustainable costs; Uber set a hard cap of $1,500 on monthly employee AI programming tool consumption; Starbucks announced a halt to AI tool deployment in North America just nine months after implementation, citing reliability issues.

These cases point to a core problem: there is currently no clear positive correlation between AI spending and AI returns. Alberto Romero believes similar stories will continue to emerge in the future because this dilemma is universal and not just a special circumstance for individual companies. He also specifically notes that the issue of hallucinations in AI models has yet to be solved, which makes him skeptical about whether AI can withstand the test of time like the internet or electricity.

Whose futures are deeply tied to AI?

Alberto Romero presents a paradox in his article that is somewhat ironic: the founders and early shareholders of AI giants, regardless of whether AI ultimately succeeds or fails, have already secured victory— success brings both fame and fortune as drivers of a technological revolution, while failure has already been cashed out through the IPO. The ones whose fates are deeply tied to the prospects of AI are, ironically, ordinary people.

He also acknowledges that if AI technology ultimately delivers on its promises, the IPO will extend what were originally private gains into public benefits, which is a normal mechanism for capital markets to promote social prosperity. However, he is not optimistic about this, as the question remains open: whether a technology with questionable reliability and an unresolved hallucination issue can truly pass the test of history.

Alberto Romero also expresses that he does not believe this is a premeditated conspiracy. In his view, the information bubble in Silicon Valley is sufficiently closed, and those founders who believe in “achieving superintelligence by 2027” likely hold their beliefs sincerely. But he emphasizes that subjective intentions do not change objective outcomes— regardless of motivations, the transfer of risk is genuinely occurring.

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。