Anthropic’s IPO filing shows a fast-growing AI business built to keep customers inside Claude

Anthropic has filed for a Nasdaq initial public offering, and the paperwork shows a company racing ahead on enterprise demand even as it posts enormous losses and leans harder into a tightly integrated Claude ecosystem. The S-1, filed Monday, lays out a business that says it is growing quickly, but is also trying to make Claude harder for large customers to leave.

That combination is what makes this filing more than a routine IPO step. According to TechCrunch, WIRED and Hindustan Times, Anthropic’s disclosure shows a company turning its flagship model, developer tools and workplace apps into a single commercial stack while raising the stakes on customer concentration, cloud dependence and the economics of frontier AI.

What Anthropic disclosed in its IPO filing

The filing makes clear that Anthropic wants to reach public markets on the strength of its enterprise traction. TechCrunch reported that as of Sept. 30, more than 11,900 customers were spending at least $100,000 annually on Anthropic products and services. More than 1,900 were spending at least $1 million, and about 250 were spending $10 million or more a year.

Those are not casual software subscriptions. They suggest that Anthropic’s products have moved deeply into major company workflows, with the company’s own prospectus portraying Claude as a layered offering: models, a developer platform and applications that reinforce one another. In the company’s framing, gains at one layer lift the others, creating a commercial flywheel rather than a single product sale.

The filing also shows that customers are increasingly adopting both the Claude platform and Anthropic’s first-party app. TechCrunch reported that more than half of Anthropic’s $100,000-plus annual customers now use both, up from roughly 30% at the end of last year. That is a meaningful change in a short time, and it points to a stronger grip on customer workflows than a model-only relationship would allow.

Anthropic also disclosed unusually strong expansion among existing users. Net revenue retention was around 900% at the end of Q3, according to TechCrunch. In plain terms, that means customers already inside the ecosystem are spending far more over time than they were when they first signed up, a sign that usage is broadening quickly across teams and products.

Why it matters

Anthropic’s filing matters because it offers a rare look at how one of the most prominent AI companies is trying to convert technical momentum into durable revenue. The market has spent two years asking whether generative AI businesses can become financially defensible. Anthropic’s answer appears to be yes, at least for now: by embedding Claude into enterprise systems, coding tools and workplace apps, it is trying to increase the cost of leaving.

That strategy has implications beyond Anthropic. If the company can make its integrated stack indispensable, it could become a model for how AI vendors move from raw-model access to recurring enterprise infrastructure. If it cannot, the S-1 shows how expensive that effort can be when scale comes with soaring compute bills, partner dependence and large losses.

For customers, the disclosure suggests that Anthropic’s products are becoming more deeply woven into business operations. For investors, it raises the central question every AI IPO will face: how much of the growth is durable demand, and how much is the result of rapid spending by a narrow base of customers at the frontier of adoption?

For cloud providers, the filing is also notable because Anthropic’s distribution model depends on firms that are both partners and potential rivals. The company says its joint go-to-market with Amazon Web Services, Google Cloud and Microsoft Azure helps it avoid channel conflict and set pricing. But as TechCrunch noted, those same cloud companies may also become competitors or affect access to compute, making the partnership structure strategically useful but not risk-free.

How big is Anthropic’s enterprise business?

Very large, if the filing is taken at face value. The clearest sign is the number of high-spending customers, which places Anthropic among the most commercially established AI companies in the enterprise market. The company says its largest accounts are expanding rapidly, not merely testing the software.

TechCrunch also highlighted a striking detail: the highest-spending 5% of Claude Enterprise customers using Claude Code spent about 20 times the monthly average of non-code enterprise users. That suggests coding products are doing more than generating buzz. They may be a major monetization engine, particularly for customers using Claude in production environments rather than as an assistant for occasional tasks.

The first-party app adoption figure supports the same story. More than half of $100,000-plus customers now use both the platform and the app, up from about 30% at the end of last year. That shift suggests Anthropic is increasingly successful at turning a model sale into a broader software relationship.

By the numbers

• 11,900+ customers spending at least $100,000 annually as of Sept. 30
• 1,900+ customers spending at least $1 million annually as of Sept. 30
• About 250 customers spending $10 million or more annually as of Sept. 30
• 900% net revenue retention at the end of Q3
• 20x higher spend among the top 5% of Claude Enterprise customers using Claude Code versus the average non-code enterprise user

Background: how Anthropic got here

Anthropic was founded in 2021 by seven former OpenAI employees, according to WIRED. It entered a market already being reshaped by large language models, then introduced Claude in 2023, months after ChatGPT made generative AI a mainstream business story.

Since then, Anthropic has positioned itself as a major rival in foundation models and enterprise AI, competing with OpenAI, Google and others. Its public pitch has increasingly focused on reliability, safety and enterprise utility, not just model size or consumer popularity. That positioning matters because the company is now asking public investors to back a business that is both fast-growing and deeply capital-intensive.

The S-1 also follows a broader pattern in the AI sector. Public filings are becoming the place where hype meets accounting. Private-market valuations can tell a dramatic growth story, but the registration statement forces companies to spell out customer concentration, revenue quality, operating costs and risk factors in language regulators will accept.

OpenAI, Anthropic’s closest and most visible rival, had confidentially filed for an IPO earlier but later said it would not go public this year, according to WIRED. That makes Anthropic’s move especially important: it may become one of the first major AI labs to test public markets at a scale that lets investors scrutinize the business in detail.

What the loss figures say about the AI business model

The revenue numbers are eye-catching, but the losses are even more revealing. WIRED reported that Anthropic’s revenue grew to more than $16 billion from about $4.5 billion in 2025, while the company’s first-half 2026 net loss reached $80.7 billion, compared with $13 billion a year earlier. Hindustan Times separately reported a nearly $42 billion net loss in 2025.

Those figures differ across outlets, and the differences matter. TechCrunch’s filing-based revenue number for the first half of 2026, $16.3 billion, is much lower than the annualized impression some readers might infer from other reports, while WIRED’s and Hindustan Times’ loss figures point to a business whose spending is still wildly out of sync with its top line.

Even allowing for differences in how outlets framed the figures, the overall picture is unmistakable: Anthropic is scaling revenue fast, but the cost of frontier AI remains enormous. Hindustan Times reported projected costs of $518 billion over the next few years, underscoring how much capital the sector may continue to consume if model performance keeps improving and usage keeps expanding.

That economic reality helps explain why public markets matter so much now. An IPO gives Anthropic access to a larger pool of capital, but it also subjects the company to a harsher form of scrutiny. Investors will want to know not only how quickly Claude is growing, but whether the company can ever grow into its cost structure.

How Anthropic says it keeps customers close

The filing’s most important strategic detail may be less about the IPO itself and more about customer retention. Anthropic is not just selling access to Claude. It is building a commercial system that links models, a developer platform, first-party apps and partner channels.

According to TechCrunch, the company says its joint go-to-market arrangement with AWS, Google Cloud and Microsoft Azure allows it to avoid channel conflict and set pricing. In practice, that means Anthropic can distribute through the biggest cloud vendors without having to choose one exclusive route to market.

It also means Anthropic’s growth is tethered to outside firms that are powerful enough to shape its business. Cloud partners can be helpful on distribution and infrastructure, but they can also compete with Anthropic in adjacent AI products or influence the terms on which compute is available. That creates both scale and fragility.

The lock-in strategy, in plain language

1. Sell Claude to enterprises as a foundational model.
2. Expand use into developer tools like Claude Code.
3. Attach first-party apps that deepen everyday use.
4. Use cloud partners to reach customers without forcing a single sales path.
5. Encourage expansion so existing accounts spend far more over time.

The prospectus also flags customer concentration as a risk. If large buyers cut spending, results could be hit quickly. That is the downside of a business with a relatively small number of very large accounts: growth can be spectacular, but so can the damage if the biggest customers slow down.

What risks did the filing highlight?

The risks are not subtle. WIRED reported that the prospectus warns AI could create catastrophic or existential risks and could be abused for cyberattacks or biological weapons. That language places Anthropic among the few major AI companies willing to state such threats plainly in a public offering document.

There is also a commercial risk embedded in the same filing. As customer adoption deepens, so does dependence on a small set of very large buyers. TechCrunch reported that the prospectus warns customer concentration could hurt results if those big customers reduce spending. In other words, the same enterprise momentum that makes Anthropic attractive also creates exposure.

Then there is the capital risk. AI infrastructure requires expensive compute, and the cost of scaling frontier models can rise faster than revenue. That is why Anthropic’s partner structure matters so much: access to cloud resources is not just an operational issue; it is part of the company’s ability to keep growing at all.

“The speed and depth of our enterprise customer adoption are powerful evidence of Claude’s position at the frontier of AI.”

Where the reports differ

Because this story is being pieced together from multiple outlets, the differences are worth spelling out. TechCrunch focused heavily on the S-1 filing itself, especially the customer metrics and product mix. WIRED emphasized the scale of revenue and losses, plus the safety language in the prospectus. Hindustan Times added reporting on the company’s possible IPO timing and valuation ambitions.

The outlets are broadly aligned on the core point: Anthropic is preparing for a public debut while reporting rapid enterprise adoption and eye-watering losses. Where they differ is in emphasis and in the way financial figures are framed. That is common in a fast-moving IPO story, especially when different reports are drawing from the same filing while adding outside sourcing and market context.

What happens next?

The next phase will be all about the timing, the size and the market reception. TechCrunch reported the S-1 was filed Monday, and WIRED said Anthropic plans a Nasdaq listing under the ticker ANT. But the exact offer size and final valuation have not been set out in the fact sheet, and that leaves major questions open.

Hindustan Times reported that the IPO may come after the U.S. midterm elections in November, and it also said the company could seek a valuation above $2 trillion. WIRED, citing The New York Times, said Anthropic could aim to raise more than $100 billion and target a valuation around $2 trillion. Those are ambitious figures, but they remain reported ambitions rather than confirmed deal terms.

Investors will also watch for whether future filings match the financial figures that have already leaked or been reported. The difference between revenue growth and the scale of losses will be one of the central tensions in the offering. So will customer concentration, compute access and the durability of the company’s expansion strategy.

Another question is whether Anthropic can maintain its current pace of enterprise adoption without losing pricing discipline or overextending its cloud partnerships. If the company can keep converting large customers into multi-product users, its commercial case gets stronger. If not, the S-1 may be remembered less as a launch pad and more as a warning about the cost of scaling frontier AI.

Why the filing is a signal for the broader AI market

Anthropic’s IPO filing is not only about one company. It is a signal that the AI market is moving from pure experimentation into the discipline of public-market reporting. In that sense, the S-1 is a milestone for the whole sector.

For enterprise software buyers, it shows that model vendors are now trying to become full-stack providers, not just API suppliers. For rivals, it shows how quickly a company can build a sales base if it lands deep inside business workflows. For regulators and policy makers, it shows that frontier AI companies are preparing to raise public money while openly acknowledging catastrophic risks.

And for investors, the filing sets up a simple but difficult question: does Anthropic represent the future shape of AI software, or the most expensive version of it so far?

Key points to watch

• Final Nasdaq IPO timing and whether the company lists in November, as some reports suggest.
• Whether Anthropic files updated numbers on revenue, losses and customer concentration.
• How the company prices the offering and what valuation it ultimately seeks.
• Whether cloud partners remain supportive as Anthropic grows larger and more strategic.
• Whether enterprise adoption continues to broaden beyond the biggest accounts.

Anthropic’s public filing shows a company with real momentum, real customers and real commercial leverage. It also shows a business still powered by extraordinary spending and extraordinary risk. That tension will define the IPO story from here.