[DigitalToday reporter Chi-gyu Hwang] Anthropic on Oct. 26 (local time) released its S-1 registration statement for an initial public offering (IPO).

With the S-1 now public, investors can review Anthropic’s business performance in specific figures rather than relying on accounts from investors or anonymous sources. Reports dissecting the filing have followed from outlets including The Wall Street Journal, The New York Times, The Information and TechCrunch.

Anthropic’s listing is likely to take place next month. The Information reported Anthropic will begin its investor roadshow in two weeks and is aiming for a Nasdaq listing in the week of Nov. 16. The ticker symbol is ANT. The New York Times reported its valuation could be as high as $2 trillion, potentially making it the biggest IPO on record and surpassing SpaceX. The Information estimated Anthropic’s valuation at at least $1.5 trillion.

Anthropic’s revenue in the first half of this year was $16.3 billion, up 14-fold from a year earlier. Second-quarter revenue alone was $11.6 billion, 2.5 times its revenue for all of last year.

Net loss in the first half was $80.7 billion, mostly due to a one-off $72.4 billion cost related to convertible bonds. Convertible bonds are debt that can later convert into shares, and the accounting expense rises as company value increases. Excluding that, the net loss was about $8.4 billion.

In the second quarter, it swung to profit with operating income of $974 million. Computing costs as a share of revenue fell to 48 percent in the second quarter from 251 percent in the first quarter of last year. Operating cash flow for the first half also turned positive at $3.9 billion.

As of end-June, cash and cash equivalents stood at $112.4 billion, about 80 percent of the $141.1 billion it has raised since its founding. The Wall Street Journal reported its small second-quarter profit contrasted with OpenAI, which posted deeper losses.

The Wall Street Journal reported that in its 400-page S-1, Anthropic effectively put the market it can target at $40 trillion, citing the economic value created by knowledge workers worldwide in 2025. That is larger than the $28.5 trillion figure SpaceX presented for its IPO.

A concern is that the money it will have to spend in the future is far greater than the money it earns now. The New York Times’ own tally showed the total value of contracts Anthropic has committed to for data centers, chips and computing power over the next 10 years is as high as $807 billion. Many are non-cancellable contracts. Even raising more than $100 billion through an IPO may not be enough, and it may need to seek additional funding earlier than the company expects.

The Information provided more specific details on Anthropic’s computing power contracts with multiple companies.

Anthropic signed computing power contracts worth $114.5 billion with Google Cloud, $110.0 billion with Amazon Web Services (AWS), $44.9 billion with Microsoft and $120.5 billion with SpaceX.

It also has a $161.0 billion contract to lease Google chips for five years through Broadcom. Its cloud commitments alone total $657.0 billion, of which $417.0 billion is non-cancellable. The Wall Street Journal reported the contract with SpaceX allows either side to cancel with 90 days’ notice.

Anthropic plans not only to rent cloud services but also to lease chips directly and deploy them in its own dedicated data centers. The Wall Street Journal reported rival OpenAI also raised its projected computing spending through 2030 to about $750 billion this year.

Dependence on partners could be a burden. Some partners supplying Anthropic with computing resources are also its investors and customers, and are also competitors. The New York Times reported Anthropic warned investors that these intertwined relationships could develop into conflicts of interest and disrupt its ability to secure computing power.

Its revenue structure is also a variable. Over the nine months from January to September this year, its top five customers accounted for 25 percent of revenue. In the first half, 48 percent of revenue came through cloud companies, and distribution fees paid to those companies amounted to 17 percent of revenue.

Its governance structure is also in focus. Anthropic offers shares in five classes. Seven co-founders hold special shares, giving them 50.1 percent of voting rights on key matters.

The Long-Term Benefit Trust (LTBT), established in 2023 to advise Anthropic on safety issues, has the right to appoint a majority of the board. But even if the LTBT opposes it, the trust’s authority can be removed if founders, employees and ordinary shareholders agree. The Information reported this could raise concerns not only among those who emphasize safety but also among those worried about power being concentrated in the founders.

Restrictions on share sales (lockups) also apply. Some insiders other than the co-founders can sell up to 10 percent of their holdings from the first trading day. It is similar to SpaceX and Cerebras, but is seen as far more complex.

It was also notable that the company warned that its own AI could be dangerous. In the S-1, Anthropic included a warning that advanced AI could pose catastrophic or existential risks to humanity. It also warned that its support for AI regulation could prompt political backlash or excessive regulation.

The current IPO market is not favorable. The New York Times reported that smart ring maker Oura and SoftBank affiliate SB Energy recently postponed listings, and that doubts are growing over whether huge investments can be recovered if demand for AI cools. Anthropic’s S-1 also excluded third-quarter results. Lightshed Partners analyst Brandon Ross (Brandon Ross) said, "Investors are looking at the latest performance to judge whether Anthropic can handle its spending commitments and grow into its valuation."