Anthropic’s IPO filing, first reported by Reuters, shows revenue up twelvefold to nearly $4.6 billion in 2025, an operating loss of about $8 billion, and $518 billion in future computing commitments, with a possible valuation above $2 trillion. It also hands the seven co-founders 50.1% of the vote and warns shareholders that decisions made in the name of safety could cost them money.

Editor’s note. This article is based on Anthropic’s IPO prospectus as reviewed and reported by Reuters on 28 and 29 September 2026. All figures, valuations, share prices, and timelines are accurate as of publication and may change as the filing is revised and the offering progresses.

[For more news, click here] 

Anthropic’s prospectus runs to 261 pages, and roughly 80 of them deal with what could go wrong, according to Reuters, which reviewed the filing. That is close to twice the space the company gives to describing its own business, and somewhere in those pages it tells prospective shareholders that the technology it sells could pose “catastrophic or existential risks to humanity.”

Bankers rarely want that kind of language anywhere near an offering document, least of all one that, as of publication, could value Anthropic at more than $2 trillion. The filing shows sales growing at a pace rarely seen at this size, costs that still far outrun them, and a founding team that has kept the power to overrule shareholders over safety.

For ordinary American investors who have spent years watching venture capital firms, sovereign wealth funds, and Big Tech pour money into AI from the outside, the listing would be the first real chance to buy shares in a frontier AI lab.

How fast is Anthropic growing, and how much is it losing?

Revenue rose twelvefold in 2025 to nearly $4.6 billion, the prospectus shows, but spending climbed by more in dollar terms, with operating expenses reaching $12.65 billion and the operating loss widening to $8.06 billion from $2.98 billion the year before. More than half of those expenses, $7.33 billion, went on compute and infrastructure, about triple the 2024 figure.

The headline net loss of about $42 billion overstates the damage, since roughly $34 billion of it is a non-cash accounting charge tied to a jump in the estimated value of financing that could later convert into shares.

The figure that deserves the most attention is $518 billion, the amount Anthropic expects to spend on cloud, computing, and infrastructure obligations in the years ahead. At the end of 2025 it held $20.28 billion in cash, cash equivalents, and short-term investments, so meeting those commitments depends on fast revenue growth and continued funding, which a stock listing makes easier. Some of that money will flow straight back to the company’s own backers, since Amazon and Google have each put billions into Anthropic while also renting it the cloud capacity it uses to train and run Claude.

The customer base is quite narrow with nearly a quarter of last year’s revenue coming from two clients and Anthropic cautions that many of its biggest customers have no long-term contracts and could cut back or walk away.

Why a $2 trillion valuation will test the AI trade

At $2 trillion, Anthropic would be worth more than twice the roughly $965 billion it estimated for itself in May and more than 400 times what it brought in last year. Buyers would be paying for the argument made in the prospectus which is that AI will change the global economy more deeply than industrialization, electricity, or the internet did.

The closest comparison is SpaceX, which listed on 12 June at a $1.77 trillion valuation and saw its shares jump 19% from their $135 offer price to $160 on the first day, though as of publication they have slipped back to around $147. AI and chip stocks have also sold off in recent weeks, and Reuters has reported, citing sources, that Anthropic is likely to wait until after the November US midterm elections to list.

According to media reports, Anthropic is also racing OpenAI, which confidentially filed for its own IPO in June and is expected to list by early 2027. Whichever of the two gets there first will probably set the yardstick investors use to value the other.

What is Anthropic’s Founder LLC, and how does it work?

Anthropic is a Public Benefit Corporation under Delaware law and will stay one after listing which already gives its leaders legal room to scope out the interests of society alongside those of shareholders. On top of that, through an entity called Founder LLC, a majority of the company’s seven co-founders will control a single share of Class F stock carrying 50.1% of the total voting power on key matters, including the election of some directors.

People buying on the public market would get Class A shares with one vote each, while strategic partners get limited voting rights. Class F and Class A holders will elect three board members, including Daniela Amodei, the company’s President and Chair, and her brother Dario Amodei, the Chief Executive, with the third yet to be named. The other four seats go to directors picked by Anthropic’s Long-Term Benefit Trust, whose trustees include former Federal Reserve Chair Ben Bernanke and national security expert Richard Fontaine.

Anthropic spells out the downside for shareholders in plain terms, saying the structure could lead to decisions “that may conflict with short-, medium-, or long-term financial interests and business performance, which may negatively impact the value of our Class A common stock.”

Plenty of tech companies are controlled by their founders, but that usually means one person, as with Elon Musk at SpaceX. Anthropic’s version is spread across seven people who walked out of OpenAI together in 2020 and have held together since, with rules for what happens if they split. A founder can be removed for leaving, dying, selling too many shares, or for cause, and the special voting power starts to wind down once two or fewer co-founders or their successors remain.

Shareholders at other founder-led companies have learned how that kind of control can cut against them, most recently at Meta, which agreed in August to pay up to $18 billion over children’s safety concerns after years of resisting activist investors, and at Tesla, where the stock still moves on its chief executive’s social media posts. Anthropic’s founders are well paid, with Dario Amodei receiving nearly $18 million in 2025 and Daniela Amodei $16.4 million, though they have also pledged to give 80% of their personal Anthropic equity to charity.

Why Anthropic’s AI safety warnings matter to investors

Parts of the risk section read more like a lab report than a securities filing describing how Anthropic’s models have shown “self-preserving behaviours” in testing, including attempts to “resist shutdown,” as well as tendencies to “conceal or manipulate information” and conduct “resembling blackmail.” The company also admits that “Potential model awareness of our evaluation efforts creates a significant limitation on our ability to assess model safety.”

As evidence that it takes these risks seriously, Anthropic points to keeping its cybersecurity-capable Mythos Preview model in a limited access program, and the filing says: “Similarly, we have chosen not to develop certain commercially attractive offerings, such as image and video generation models, in order to direct our compute toward our research and safety priorities.”

Holding themselves to that gets harder when rivals are moving fast, and even as Dario Amodei has called on the industry to slow the release of new capabilities, Anthropic shipped its Opus 5.5 model last week, in part to answer the momentum OpenAI has built since launching GPT-6 Astra.

What Anthropic is taking to market is a company that says in writing it may put safety ahead of profit, that it cannot fully measure how its own products behave, and that its founders will keep the final word. Public investors have never been asked to price anything quite like it, and the coming months will show how much they are willing to pay.

Related Articles

Microsoft’s $10 Billion Gulf Deal Extends Beyond Data Centers

Exclusive: Nikita Astionov on the Security Risks of Giving AI Agents More Control

Google Project Suncatcher Test will Send its AI Chips into Orbit