AI 泡沫疑慮籠罩 Anthropic 的 IPO 時刻

Move over, SpaceX: There may be a new record-breaking IPO about to come to market. Anthropic is set to make its debut as a public company in November after filing S-1 paperwork Monday.

The company, known for its Claude AI products, first filed confidentially in June. While SpaceX beat its AI industry peers to the punch, investors have been eyeing the IPOs of Anthropic and OpenAI for months. While OpenAI has eased its ambition to go to market in 2026, Anthropic has moved forward, even as investors increasingly question how the AI industry can generate the trillions of dollars needed to pay down its debts.

With the so-called safety-first AI giant’s prospectus now publicly available, it’s clear this isn’t a standard IPO filing. Nearly a third of the document elaborates on risk factors, warning that Anthropic’s “development of highly advanced models, platforms and applications and expansion of use cases could further increase the risk that our models cause harm.” Risks are a major part of any company’s disclosures to investors — if they’re not disclosed, investors can sue.

More from CNET

AI company CEOs, like Anthropic’s Dario Amodei, are consistently open about the risks posed by their products, up to and including human extinction. While that might seem counterintuitive, doomsday talk is a key to boosting the marketing hype around AI.

Hungry investors are likely to throw in for Anthropic regardless of the financial risks, after years of AI company investment being available mostly to private venture capital firms. Frontier AI labs continue to race toward the creation of new AI models, spurred on by vague promises of future AGI, or artificial general intelligence — a theoretical moment when the technology allegedly matches or surpasses human capabilities.

Yet Anthropic may have missed its window for maximum enthusiasm. In the last few months, reports of poorly controlled AI models in testing environments hacking into external websites have raised red flags and pointed to the need for more regulation and slower growth. The ongoing public backlash against data centers and Big Tech has dampened the hype around AI from where it stood just a season ago.

What’s in Anthropic’s IPO filing?

Anthropic is looking to execute its IPO the week of Nov. 16, according to a report by The Wall Street Journal. The company is looking to trade on the Nasdaq under the ticker ANT, and it’s expected that this market debut will raise more than $100 billion in liquidity for Anthropic.

The S-1 filing has revealed startling information about Anthropic’s financial position. In the first half of 2026, the company generated around $16.3 billion in revenue, a massive increase from the $1.2 billion reported in the same period just one year earlier. Yet Anthropic also disclosed an $80.7 billion net operating loss during the first half of the year, “a figure that largely reflects an accounting adjustment related to financing agreements,” according to The Wall Street Journal.

The shift to a November IPO may mean Anthropic will be obligated to share its third-quarter financial information with potential investors up front.

There are massive gaps in Anthropic’s filing that need to be ironed out over the next few weeks. Notably, no information has been shared about the company’s total valuation or how many shares it plans to sell, though it’s reportedly eyeing a valuation of up to $2 trillion. According to The Wall Street Journal, those details “likely won’t come for another two weeks or so, when Anthropic files updated paperwork after gauging investor feedback.”

Any valuation in trillions is an absurd amount of money that’s hard to visualize. If you earned one dollar every second, it would take you more than 31,700 years to earn $1 trillion.

Does Anthropic’s speculated valuation of nearly double that amount actually make any sense? Probably not, considering it’s based on revenue forecasting that projects growth rates that have never been achieved by any other company, ever. Plus, considering Anthropic’s previous internal valuation of roughly $965 billion, that’s far from the $2 trillion mark.

The IPO valuation is based on Anthropic’s unprecedented future revenue estimate, which projects it will generate between $190 and $ 200 billion annually by 2028. This is despite its S-1 filing acknowledging a “limited operating history in an evolving and nascent industry, which makes it difficult to evaluate [its] future prospects and may increase the risk that [it] will not be successful.”

Regardless of how ridiculous that may sound, this accounting approach has precedent in the tech world (especially among AI companies). SpaceX’s IPO hinged its valuation on projected 2029 revenue.

Just like Anthropic, SpaceX’s eyebrow-raising revenue numbers secured ample funding for the company, despite never having turned a profit throughout its existence. At the very least, the Elon Musk-owned company has a robust set of satellite connectivity and aerospace-related assets on its balance sheet, tangible assets Anthropic doesn’t have.

Moreover, Anthropic played fast and loose with the disclosure of important tidbits about its expenses in its IPO filing. The company’s reputation banks on safety being the appealing market alternative, but Anthropic’s documentation said that returns on its guardrail investments remain unclear.

Without sharing exact spending figures, the company described its implementation of safety features as “resource-intensive” and said it is split between safety measures and computing power. It seems Anthropic is preparing investors to see some ugly spending now and in the future.

It’s also worth noting that Anthropic’s projected revenue depends almost entirely on enterprise and power usage of Claude models at a time when businesses are feeling burned by token costs and are pulling back on AI usage.

The IPO filing comes just days after Apollo Chief Economist Torsten Slok observed that AI has “delivered zero financial payoff to corporate America outside of tech,” with rising profit margins among S&P 500 firms confined to the so-called Magnificent 7 Big Tech companies. 

There’s no telling what could happen to the AI spending bubble after the dust settles around this IPO. We know it’ll generate a ton of paper wealth for Anthropic insiders, certain venture capitalists and other large investors, but the company is increasingly competing on price with OpenAI and cheaper open-model developers at a time when major AI firms already report spotty profitability.

According to AI industry critic Ed Zitron, the math behind this actually makes explaining the bubble “super simple.”

“Hundreds of billions of dollars are being invested to make single-digit billions of dollars maybe, some day, if AI data centers actually get built at scale and OpenAI and Anthropic can afford to pay for their compute,” Zitron wrote.

Compounding concerns for the biggest AI companies include the exorbitant price of data center infrastructure necessary to process and store LLM queries and the proliferation of free open-source models that serve as viable alternatives to Claude for certain clientele.

Looking at the industry’s power players reveals how shaky the financial foundations are now. During a Newsweek interview, Zitron pointed out that OpenAI has committed to spending $600 billion on compute by 2030, with payment obligations starting as soon as 2027.

“Everything is relying on a system that doesn’t exist anymore,” Zitron said.

If Anthropic and OpenAI can’t turn a profit now, does it make any sense for these AI giants to race each other to the bottom on token pricing? Probably not, which might be why now is the time to sell off pieces of the company to you, the retail investor, to secure golden parachutes for all the insiders who have built this house of cards.

In truth, the $2 trillion question really isn’t whether Anthropic can deliver on the wild revenue goals set by its analysts. It’s how much the American public will be hurting if and when the industry’s circular financing dries up, and the remaining money gets pissed away with no profits to show for it.

Public sentiment toward AI is turning sour

The wider public isn’t backing the AI industry, financially or otherwise. As LLMs have proliferated and AI has been integrated into more facets of everyday life over the past few years, public sentiment has shifted toward overwhelming negativity.

Younger generations are turning against the very concept of LLMs, as Anthropic is allegedly hiring fewer junior technical workers than ever before. Amodei regularly makes off-the-cuff statements that AI company executives need to “stop sugarcoating” a future in which he envisions agentic models wiping out 50% of entry-level positions. It’s hard to get excited about a future that seemingly doesn’t have a place for you.

The company frames AI as a product that will usher in a new industrial revolution or cause irreversible harm, depending on the way it’s developed, claims that generate hype and fear in equal measure. “Concerns have been raised both within our organization and by other researchers, regulators, policymakers, and industry participants that sufficiently advanced AI systems could, if not developed, deployed, and operated with appropriate safeguards, oversight, and security measures, pose catastrophic or existential risks to humanity, particularly if such systems were misused, replicated, weaponized, or allowed to operate with a high degree of autonomy,” Anthropic said in the filing.

In September, Anthropic researcher Evan Hubinger stated publicly that he estimated a more than 10% chance that AI could kill humans within the next decade. That tidbit came after similar viral statements by ex-Anthropic researcher Jacob Coxon.

Such assertions are inherently unverifiable, but serve as implicit marketing for Anthropic. Still, such comments are a lightning rod for negative public reactions, which Amodei isn’t blind to. He acknowledged in an X post that there’s “fundamentally a crisis of trust” between the public and AI companies.

The Anthropic CEO routinely makes statements designed to placate the public, issuing weak pleas about slowing down the industry’s development and being more responsible — while still being locked in a race with other AI developers to develop and release powerful models. In September, he published a blog calling on his peers to “slow the pace at which [they] improve the capabilities of AI models.” Within a week of this blog post, Anthropic had launched a new, more powerful Opus model.

Much of Amodei’s talk about slowing down LLM development until the necessary guardrails are in place will likely go out the window after the IPO. Sure, Anthropic’s IPO document says the company “believes building reliable, trustworthy and secure AI systems” will be rewarded by the market. But once it is beholden to the whims of its investors, Anthropic will be expected to produce unending quarterly gains.

In its filing, Anthropic says it needs to stick to a “continuous and overlapping cadence” of model releases in order to remain “at the frontier of AI development.” The infinite-growth mindset doesn’t leave room for public-safety guardrails. If anything, we should expect development of new Claude models to accelerate post-IPO — and public distrust of the company (and the AI industry as a whole) to grow alongside them.