The SEC Should Block Anthropic's IPO to Protect the Public

Matt Scherer is co-author of the latest study Brace for Impact: Protecting group and the economics whenever the AI bubble bursts.

Things are rarely what they appear whenever it comes to the finances of the AI industry. The AI ecosystem’s circular endeavor dealings, extended use of financial engineering, and alternatively murky finances create it difficult for investors, regulators, or workers preservation for withdrawal to get an exact financial image of AI companies.

That is particularly true of Anthropic, which submitted the registration statement for its first community offering (IPO) final week onward of a planned November inventory market debut.

The company’s topline financial figures appearance fairly notable for a startup, alongside revenues that appear to have grown quickly complete the way of 2026 and a fairly fit balance panel — on the surface, that is.

Dig a small deeper, however, and a far bleaker image emerges. Footnotes to the company’s most latest financial disclosures revealed nearly $900 billion in spending commitments, most of them stretching into the 2030s. The company’s financial results for 2026 have not been subjected to ceremonial auditing, which makes them less-than-reliable, but the third-quarter results propose that the company’s income growth slowed to a crawl complete the summer.

Anthropic volition have to reverse that trend and prolong bonzer income growth to encounter its huge obligations. But the financial pressures that are squeezing Anthropic — and, indeed, the AI industry as a entire — volition create it exceptionally difficult for the foremost example developer to do so. If the business goes public, workers, retailing investors, and retirees volition increasingly be remaining holding the bag whenever Anthropic’s bills arrive due.

The financial mountain that Anthropic must ascend is unprecedented for a business making its inventory market debut. While Anthropic has an notable $96 milliard in prosperity reserves, its amended registration statement reveals at smallest $582 milliard in non-cancelable spending commitments to fair five companies (Amazon, AMD, Broadcom, Google, and Microsoft) to acquire admission to the huge computational power required to train and run Anthropic’s generative AI models.

Even although they do not appear on Anthropic’s balance sheet, those obligations are efficiently debts in a lawful and applicable sense: they are payments that Anthropic is contractually required to create in the future, despite of whether the business can oversee to do so.

Moreover, it is not apparent how much of the remaining $300 milliard or so of the $895 milliard in contractual commitments volition ultimately district on Anthropic’s balance sheet. With the business continuing to negotiate further enormous infrastructure and compute contracts, Anthropic may fine end the twelvemonth alongside a trillion dollars or additional in contractual commitments.

Generating the revenues necessary to encounter those obligations would be, to say the least, fairly difficult for any company. But the margins of AI example developers are being squeezed by multiple forces that volition create that project extraordinarily challenging for Anthropic.

First, it isn’t apparent that there is nearly adequate petition for generative AI products and services to validate Anthropic’s spending. Multiple studies, including from organizations notable for their bullishness on AI, propose LLMs are not accelerating productivity or otherwise generating affirmative returns on investment for most businesses. Worryingly for enterprise-focused Anthropic, one study established that the figure of Americans using AI at activity fell significantly between the summers of 2025 (46%) and 2026 (33%).

Anthropic’s spending commitments propose that the business believes it volition be capable to convince businesses to pay for ever-larger and additional costly models. But there are expanding signs that most businesses are not curious in paying for additional mighty and costly models. When Anthropic released its much-hyped Fable 5 frontier example this summer, its business reception was tepid.

Even if a forthcoming Anthropic example captivates businesses, the business would run into a second obstacle: the AI industry’s frenzied competition. Anthropic is part of a crowded Silicon Valley ecosystem anywhere many well-heeled corporations (including Google, Meta, Microsoft, OpenAI, and SpaceX, alongside alongside Anthropic) recommendation generative AI models that are mostly indistinguishable in their norm and capabilities. As a result, AI companies can quickly recommendation competing products whenever a competitor identifies a use case that shows equal a clue of business potential.

In specified an environment, companies can rarely oversee to lift prices since doing so volition immediate customers merely to toggle to a rival. The hurry to catch customers onward of their IPOs led Anthropic and OpenAI to initiate a price war this summer that remains ongoing. Competition from Chinese companies offering open-weights models puts further downward pressure on prices, as does the expanding preparedness of models that can run locally on phones and laptops — no costly data centers required.

That leads to the third problem: the extraordinarily elevated expenses of the chips and data centers needed to train and run AI models, which put mighty and uninterrupted upward force on Anthropic’s costs. Between the chips, servers, networking equipment, power and cooling infrastructure, building costs, and another expenses, a 1 gigawatt (GW) data center expenses an estimated $31 billion, and the prices for several components are soaring.

That is for 1 GW. Anthropic has committed to 15 GW of data center capability in the coming years and the business estimates it volition ultimately need 20-25 GW, which explains the company’s half-trillion-plus dollars in spending commitments.

And that is fair the disbursal of the bodily infrastructure. Once a data center is up and running, training AI models entails using tens of thousands of chips operating in parallel that necessitate enormous amounts of energy to power, run hot, and need to be kept cool. Moreover, all period person prompts a example to answer a question, create an image, or execute a task, additional computational power, energy, and cooling are needed to procedure the query and create a reply or complete the task.

Finally, there are external factors that are delaying data center building and piling on additional costs. The fierce and still-rising backlash to AI in general and data centers in particular throughout the nation has scuttled a figure of data center projects and delayed many others. Every postpone limits the computational capability Anthropic is counting on to create the income it volition desperately need, and rising expenses because of the Iran War and supply-chain challenges could run Anthropic’s operational expenses motionless higher.

Anthropic volition additional pad its prosperity reserves if its IPO is successful, but the ruthless economics of its endeavor are not apt to enhance anytime soon. Even if petition picks up significantly, serious competition volition use unrelenting downward force on prices equal as elevated expenses and asset constraints compression it from the contrary direction. That is a deadly mix for Anthropic in ray of its slowing income growth and enormous off-balance-sheet debts, a huge and immovable fixed cost.

If Anthropic were to go community now, it would heap additional hazard onto a inventory market that is already historically overvalued and disproportionately concentrated in a fistful of tightly interconnected mega-cap companies from the tech sector. That has already created a hazard of a historically damaging inventory market crash; never before has specified a huge amount of prosperity depended on the achievement of a small figure of companies whose fortunes are inextricably linked. Adding another huge, unprofitable tech elephantine to community exchanges would exacerbate the two of those factors and create a higher hazard of contagion if (or, rather, when) Anthropic reaches a financial precipice.

It additionally would average that an ever-larger portion of the fallout from a possibly devastating AI collision would autumn on average people. As a personal company, Anthropic’s shares are overwhelmingly owned by company insiders, tech giants, and project capitalists. Once the business goes community and shares commencement getting funneled into 401(k)s, withdrawal accounts, and indicator funds, it volition be workers, retirees, and retailing investors that are increasingly remaining holding the bag whenever the forces squeezing Anthropic’s margins commencement to bite.

The Securities and Exchange Commission (SEC) has the power to halt that from happening. The agency can postpone or halt companies from going community if it believes a company’s registration declaration is “on its visage incomplete or inaccurate in any matter respect.” That legislative tongue power as fine have been written alongside Anthropic in mind, considering that the company’s prospectus includes lone unaudited financial results for the duration anywhere it claims the top income growth, relegates its most important forthcoming obligations to the footnotes of its registration statement, and has continued to province new 11-digit financial commitments as its IPO approaches.

The SEC should block Anthropic’s inventory market debut until the business provides two successive quarters of audited financial statements, alongside all matter data concerning the company’s spending commitments disclosed. Failing to do so could have dire consequences for workers, investors, and ultimately the financial scheme and economy.