# Anthropic files for its stock market listing: AI is now trading

> Anthropic files for its Nasdaq listing under the ticker ANT: $16.3 billion in revenue, $80.7 billion net loss and $542 billion committed to compute.
> By José Maldonado · Artificial Intelligence · 2026-10-28

For two years Anthropic has been warning that the artificial intelligence it builds could become uncontrollable. This Monday it asked the public market for money to keep building it. The company behind Claude filed its S-1 form with the SEC, the [public registration](https://www.sec.gov/Archives/edgar/data/2130357/000162828026067168/anthropicpbc-sx1.htm) for the transaction, to list on the Nasdaq under the ticker ANT, and it did so with two figures that explain the business: $16.3 billion in revenue in the first half of 2026 and a net loss of $80.7 billion.

The loss is scarier than it should be. About $72.4 billion correspond to an accounting charge for the revaluation of the convertible bonds with which the company financed itself, according to the [breakdown of the accounts](https://fortune.com/2026/10/26/anthropics-s-1/), not to money that left the coffers. The operating loss narrowed to $693 million from $2.1 billion a year earlier, and operations generated $3.9 billion in cash for the first time.

That is the crux of this Anthropic IPO. The lab that presents itself as the most cautious in the sector needs capital from investors who look at the next quarter, and its prospectus acknowledges compute commitments that are more than thirty times a semester's revenue. The contradiction between declared safety and fiduciary duty stops being a forum debate and becomes a legal obligation before the US regulator.

## What the 90-page risk filing says

The price and the number of shares have not yet been decided. The prospectus arrives without a range and without a size, and those figures will appear in an amendment once the company finishes sounding out investors, foreseeably in a couple of weeks.

The document that entered the SEC system this Monday is not a piece of marketing: it is more than 90 pages of risk factors listing what can go wrong, from customer concentration to political resistance to data centers, according to the [reading of the transaction](https://www.cnbc.com/2026/10/26/anthropic-ipo-live-updates-ai-s-1-filing.html) published by CNBC. Anthropic did not hide the usual warning, that its technology can cause serious harm, and it put it in the same package that any investor signs.

The semester's figures support the friendly part of the story. Revenue went from $1.1 billion to $16.3 billion in a year, almost fifteen times more, and in the third quarter the company reported $16.8 billion to investors, 45% above the previous quarter. At that October meeting it said it was already profitable on an adjusted basis, with a margin close to 10%.

Item

First half of 2026

Revenue

$16.3B

Net loss

$80.7B

Accounting charge for convertibles

$72.39B

Operating loss

$693M

Cash generated by operations

$3.9B

Read cold, the table tells two stories at once. One is a business that multiplies its revenue fifteenfold in twelve months, something no investment bank expected to see in a five-year-old software company. The other is a company that still loses money on operations and needs to grow fast to cover what it has signed.

## The spending that decides the transaction

Anthropic spent close to $9 billion on compute and infrastructure between January and June, more than the $7.33 billion of all of 2025 and far above the $2.3 billion of the same period a year earlier. The prospectus declares commitments of at least $542 billion. For 2028 and 2031 it adds extra capacity of between $125 billion and $150 billion whose price is not yet settled.

The figure that conditions the next decade is not on the income statement. The company has promised to pay more than half a trillion dollars in compute, thirty times what it bills in six months, and much of those contracts cannot be reversed.

Counterparty

Declared commitment

Equipment leasing with Broadcom and Google TPUs

$161.2B

Google Cloud

$114.5B

Amazon

$110B

Microsoft

$44.9B

Those four items alone exceed $430 billion and do not exhaust the list: the document adds agreements with other cloud platforms and more than $20 billion in systems based on AMD. The detail matters because Google, Amazon and Microsoft are not just any suppliers, but shareholders in the company and its main sales channel.

A signed contract is paid whether the model is used or not, so every weak quarter becomes a cash problem before it becomes a reputational one. The comparison with revenue explains the haste: $16.3 billion billed in six months against commitments that exceed that figure more than thirty times.

## Who is in charge after the offering

The registration makes clear who retains control once trading begins. Anthropic was incorporated as a public benefit corporation in Delaware, a figure that obliges its directors to balance shareholder interest with its mission to develop responsible AI. The company also created a Long-Term Benefit Trust whose trustees elect the class T directors, a majority of the board, as detailed in the [text of the form](https://www.anthropic.com/news/s1-sec).

On top of that layer, the seven cofounders will control 50.1% of the vote through a single class F share with no economic rights. With that design, the company falls into Nasdaq's controlled company category and is exempt from having a majority of independent directors. The system solves the problem the lab has been pointing out for years, that of an AI that can escape the control of its creators.

That shield does not come free. If the board can ignore new shareholders in a decision about safety, the market will discount that risk in the price, and Anthropic will need the funds now discussing the valuation to pay the governance premium.

The problem is that listing changes the calculation. A company that reports results every ninety days responds to different incentives from those of a lab that only answered to its backers, and the prospectus acknowledges the risk by placing model safety among the factors that can harm the business. As already happened when it turned its caution into a market argument around [existential risk](https://observatorioblockchain.com/en/ai/anthropic-bolsa-riesgo-existencial/), the containment narrative has to coexist with a revenue forecast.

## What whoever buys ANT is buying

The prospectus allows measuring the real demand behind the story. As of September 30, 11,900 customers were spending at least $100,000 a year on Anthropic, another 1,900 exceeded a million and 250 were above ten million, while two customers concentrated close to 20% of the semester's billing. Almost half of revenue came through the cloud marketplaces of Amazon, Google and Microsoft.

Investors are discussing a valuation of $1.5 trillion, below the $2 trillion taken for granted a couple of weeks ago, and the company seeks to raise up to $100 billion, according to the [context of the offering](https://www.investing.com/news/stock-market-news/factboxanthropic-moves-closer-to-blockbuster-listing-with-us-ipo-filing-4972888) reported by Investing.com. Marketing would begin the week of November 9 and the transaction could surpass the $86 billion raised by SpaceX in June.

The competition appears in the document itself, with OpenAI and Google on the list of direct rivals. All three labs have coincided in calling for restrictions on open-weight models with a [joint proposal](https://observatorioblockchain.com/en/ai/google-openai-y-anthropic-proponen-restringir-la-ia-abierta/), a useful paradox: those competing for the most expensive capital in history also sign letters asking that the competition they do not control be limited.

The calendar matters more than the story. If the offering closes in November, Anthropic will be the first company focused on frontier models to list, and its price will set the benchmark for everything that comes after.

## Caution as an asset that is now traded

For two years, Anthropic's sales pitch was containment: models with a written constitution, safety audits and explicit warnings about the risk of its own tools. The stock market listing does not contradict that narrative, it puts it to the test. Whoever buys ANT at a valuation of one and a half trillion is not paying for caution, they are paying for market share.

What happens in the coming weeks will say something about the rest of the sector. If the market accepts that multiple with a half-year loss of $80.7 billion and $542 billion committed to compute, OpenAI will have a clear path for its own offering and the industry will have found its definitive financing model: public money for a technology its creators describe as existentially dangerous. If the price falls, the conversation about the real cost of AI will stop being a debate among analysts.
