Reading Anthropic’s S-1.
First-Half Revenue Up 14x Year on Year,
and 50.1% of the Vote in One Share.

On October 26, 2026, Anthropic, the company that makes the conversational AI “Claude”, filed a prospectus on Form S-1 with the SEC for a listing in the United States. It plans to list on Nasdaq under the ticker “ANT”. Revenue for the six months ended June 30, 2026 was $16.3 billion, 14 times the $1.2 billion of the same period a year earlier.

In the three months ended June 30 alone, revenue reached $11.6 billion, and the company posted an operating profit for the first time in the six quarters shown in the table. Meanwhile, the net loss for the six months was $80.7 billion. Of that, $72.4 billion is the change in the fair value of convertible notes and similar instruments, a loss that grows as the company’s valuation rises.

Even after the listing, holders of the listed shares (Class A) generally do not get to decide where the company goes. On matters submitted to stockholders generally, 50.1% of the voting power is fixed in one share of Class F common stock, held by a company (the Founder LLC) whose initial members are only the seven founders. Separately, four of the seven directors are elected by a trust (the LTBT) whose trustees have no financial stake in the company. The 50.1% counts in electing the other three. The S-1 itself writes that holders of the shares being listed generally will not be able to affect the outcome, “regardless of how they vote”.

After the cover comes a letter signed jointly by the seven founders. It passes on a piece of advice it says they heard from several business leaders while preparing for the IPO: “never let your company be worth a dollar more than it is actually worth”. A business whose revenue, by the company’s own account, reflects the fastest enterprise adoption of any technology in history is going public together with a governance under which shareholders generally cannot affect the outcome.

Ten questions. We lay out only what the S-1 says.

Question 1

What is this company, really?

In one line: a “research lab” that builds AI models and sells them by the amount used.

The company defines itself this way.

“

We are first and foremost a research lab—our mission is the responsible development and maintenance of advanced AI for the long-term benefit of humanity.

ANTHROPIC S-1 (OCT 2026) — PROSPECTUS SUMMARY

What it sells is a family of AI models named “Claude”. There are four lines: Mythos at the top (the generally available version is called Fable), Opus, Sonnet, and Haiku, the cheapest and fastest.

It sells them in three layers: the models, a platform for building on them, and applications. Developers build the models into their own products through the API. The applications the company makes itself include Claude Code, which writes code, and Claude Cowork, which takes on knowledge work. Claude can also be bought through the clouds of Amazon, Google and Microsoft.

The flagship is Claude Code. The company calls it the product it believes “has most concretely defined what autonomous AI work looks like”. One example in the filing: NASA engineers used Claude Code to plan the navigation commands for the Mars Perseverance rover, resulting in the first AI-planned drive on Mars.

It was founded in 2021. It launched its first commercial product, Claude, publicly in the first quarter of 2023, and revenue started there, from zero.

Right after the cover sits an eight-page letter. It is signed by the seven co-founders. It lists four things they would like any investor to understand.

• The incredible potential of AI
• The risks that must be addressed
• Anthropic’s unique approach to building a company and operating in the world
• The ways in which its behavior may differ from that of a purely profit-maximizing company

The second half of the letter divides the decisions that may run against investors’ expectations into two categories: business bets that derive from AI’s steep trajectory, and decisions that depart from profit maximization. As an example of the first, it writes that Anthropic will often make “forward investments (especially in compute) that may seem irresponsibly large at first glance”. Compute means the chips and data centers used to train and run AI.

The company’s own AI is used inside the company too. In September 2026, over 90% of the contributions to the company’s codebase were written by Claude.

It calls itself a research lab and sells AI by the amount used. In September 2026, its own AI wrote over 90% of the contributions to its codebase.

Question 2

What does it compete on?

It names five strengths. The one it calls especially significant is a “loop” in which AI helps build the next AI.

The company lists five core advantages.

• Focus on the mission
• People and culture
• Frontier research and safety research
• Its approach to compute
• Its enterprise strategy

A diagram in the summary shows two self-reinforcing cycles that, the company writes, its rapid adoption generates. An “improvement loop”, in which better models find more uses, and more use gathers more material for improvement. And an “investment loop”, in which revenue growth draws investment and investment raises capability. Three of the pillars follow, in turn.

AI helps build the next AI

Each generation of Claude is used in the work of building the next. The company calls this “closing the loop”, and writes that it believes it is one of the most significant structural advantages in frontier AI. Running experiments, generating training data, automating evaluations and safety audits, writing and reviewing code: Claude is part of that work.

Numbers come with it. The average daily volume of code contributed per engineer in the three months ended September 30, 2026 was over 45 times the average in 2024. The value of the company’s internal Claude usage at list prices grew by over 250 times from January 2026 to September 2026, and in April it first exceeded monthly personnel expense, excluding stock-based compensation.

The same section also states the risk. Using AI to help build more capable AI could pose a considerable safety risk, particularly if it sharply accelerates the pace of progress. The company’s Responsible Scaling Policy covers monitoring and mitigating the risks as it approaches what has been termed “recursive self-improvement”.