Anthropic opens its books: the IPO of a company that warns its technology may be an existential risk

⏱️ Reading time: ~5 minutes

Introduction

On Monday, October 26, 2026, Anthropic made its IPO filing public with the SEC, the U.S. securities regulator. Its shares are expected to trade under the ticker ANT. Until now, the market knew the company through private rounds, revenue estimates and executive statements. With the prospectus — the so-called S-1 — it now knows it through audited numbers.

And those numbers tell two stories at once. One is of growth rarely seen at a company this size. The other is of a company that writes, in the very document used to sell its shares, that the technology it sells may pose catastrophic or existential risks to humanity. Few public offerings have opened with that kind of warning.

The numbers nobody knew

Revenue. $16.3 billion in the first half of 2026, compared with $1.2 billion in the same period of 2025. A jump of roughly 13.5x in twelve months.

Losses. The operating loss fell to $693 million, from $2.1 billion a year earlier. The net loss of $80.7 billion looks alarming at first glance, but it comes almost entirely from non-cash accounting adjustments tied to convertible debt. For anyone reading the financials, the relevant figure is the operating one: the company still loses money, but it loses far less while growing far faster.

Customers. As of September 30, there were more than 11,900 customers spending at least $100,000 a year, more than 1,900 above $1 million and about 250 above $10 million.

Retention. Net revenue retention was around 900% in the third quarter. In plain terms: existing customers spent, on average, about nine times more. That is the kind of metric that explains the revenue — growth isn't coming only from new customers, it's coming from usage expanding inside the ones already there.

The strategy: come in through the API and stay

The prospectus describes an expansion model similar to Amazon's and Uber's. Customers start with the API, consuming the model inside their own systems, and then add the company's own products, such as Claude Code, aimed at programming. More than half of customers spending over $100,000 a year already use both the platform and at least one of Anthropic's own apps — at the end of last year that figure was about 30%.

There is also a commercial decision that will catch the eye of anyone who negotiates infrastructure contracts. Anthropic sells through AWS, Google Cloud and Microsoft Azure, but it keeps the contract with the customer and pays each cloud a fee. That lets it control pricing and avoid channel conflict with its own suppliers.

The governance that divides opinion

The control structure is the most debated point. The seven co-founders will form their own entity, the "Founder LLC", holding 50.1% of shareholder voting power. Most board members will continue to be appointed by the Long-Term Benefit Trust, an independent body created in 2023 to protect the company's safety mission.

In practice, those who buy shares will have an economic stake but little say over decisions. Anthropic presents this as a guarantee that the mission won't be traded for short-term results. Critics see a familiar design: public capital with closed control. Both readings fit in the same document.

The risks, in the company's own words

The risk section is unusually candid. Besides the warning about catastrophic or existential risks, the company acknowledges that it may not achieve or sustain profitability. It also records a structural tension: Amazon, Google and Microsoft — which supply its computing capacity — increasingly see it as a competitor and may prioritize their own needs or those of rivals they have invested in. And it flags revenue concentration: losing large customers would have a material impact.

What the prospectus does not yet include is the most awaited part: the valuation the company will seek and how much it plans to raise. Before the filing, talk was of a valuation of up to $2 trillion, versus $965 billion in its last private round.

The counterpoint from the same weekend

A day before the filing, Apollo Global Management's chief economist, Torsten Slok, published an analysis that goes straight at the buyers' thesis. Looking at S&P 500 companies outside big tech — what he calls the "S&P 493" — Slok found no productivity gain: margins and revenue per employee are flat. His conclusion is that AI has delivered zero financial payoff to companies outside the tech sector. In the line that made the rounds, AI "may be a great calculator, but in business, intelligence is mostly judgment."

This is where the two ends meet. Anthropic shows that companies are spending a lot — and more and more — on AI. Slok says that spending has yet to show up in the results of those footing the bill.

Conclusion

For anyone who has followed the technology market long enough, the script is familiar: explosive growth, shrinking losses, control concentrated in the founders' hands and an investor buying the promise before the return. What is new is the supplier stating, in the official document, that its product could be dangerous on a civilizational scale — and still, or precisely because of that, keeping control of decisions for itself.

The question Anthropic's IPO leaves open is not whether the company will grow. The numbers answer that. The question is whether the customers now spending nine times more will manage to turn that spending into results. If they do, the S-1 will look modest. If they don't, Apollo's analysis will have been the first warning.