Why Anthropic Doesn't Care About Individual Users, as Seen in Its Prospectus

The big one is finally here.

On October 26, Anthropic publicly filed its S-1 prospectus with the U.S. Securities and Exchange Commission, preparing to list on the Nasdaq Global Select Market under the ticker symbol ANT. In media reports, the company's valuation discussion range has reached $1.5 trillion to $2 trillion, while the offering price range has not yet been disclosed.

Within 24 hours, this 300-plus-page prospectus had been thoroughly picked apart by the tech and finance circles, and everyone has already seen the most eye-catching figures: revenue of $16.286 billion in the first half of 2026, compared with only $1.163 billion in the same period last year; operating profit of $974 million in the second quarter, marking the company's first operating-profitable quarter in its history; and as for the net loss of up to $80.7 billion on the books, the bulk of it is $72.388 billion in fair value changes on convertible notes, derivative liabilities and SAFEs, which are non-cash accounting charges.

But beyond these Wall Street narratives, domestic developers and everyday users have a different, split experience. In the eyes of users in mainland China, Claude is probably the AI product on the market today that bans the most accounts, has the strictest risk controls, and is the most infuriating to use.

Why does a company that has built a top-tier large model always seem so cold when dealing with ordinary individual users?

After going through the financial notes and risk disclosures in the prospectus, the answer is actually written right there in the books. In Anthropic's business landscape, the real money contributed by individual users accounts for a very small share.

01 Less than 20% individual users

To understand who a company truly cares about, the most direct way is to look at its revenue composition table. In note 2 to the financial statements in the prospectus, Anthropic splits revenue into two major blocks: usage-based revenue and subscription revenue.

The data contrast is stark. For the full year 2025, usage-based revenue was $3.796 billion, while subscription revenue was only $789 million, a ratio of roughly 83% to 17%. By the first half of 2026, usage-based revenue had surged to $13.064 billion, while subscription revenue was $3.222 billion; the former still accounted for about 80%, with subscription revenue at only about 20%.

The more critical detail lies in the composition of subscription revenue. Within that 20%, there are both individual users' Claude Pro and Claude Max and enterprise-level Team and Enterprise plans, and the prospectus does not break out individual subscriptions separately. This means that the revenue contributed by individual users paying out of pocket for memberships definitely accounts for less than 20% of Anthropic's total revenue, and may even be much lower.

So where does the 80% of usage-based revenue come from? The answer is cloud vendor channels and tens of thousands of enterprises.

Revenue sold to customers through the marketplaces of cloud vendors such as Amazon, Google and Microsoft has soared all the way: about 11% in 2023, about 32% in 2024, about 47% in 2025, and about 48% in the first half of 2026. Nearly half is supported by the distribution pipelines of the three cloud giants.

What supports this money-printing machine is a rapidly expanding group of institutional customers. As of September 30, 2026, customers contributing more than $100,000 in revenue over the past 12 months (counted by independent organization) had surged from more than 950 a year earlier to more than 11,900. The customer structure is also healthy: in the first nine months of 2026, no single customer accounted for more than 10% of revenue, and the top five customers combined accounted for about 25%.

The enterprise customer growth chart in the prospectus: customers contributing more than $100,000 in revenue over the past 12 months grew from about 20 at the end of 2023 to more than 11,900 by the end of September 2026 | Image source: Anthropic

The prospectus uses a specific term for individual paying users: "prosumers," defined as individual users who subscribe to AI tools to do knowledge work, such as researchers, developers, analysts and writers. The title sounds professional, but these people still remain marginal when it comes to core financial metrics.

The clearest expression of this attitude is net revenue retention.

As of September 30, 2026, Anthropic's net revenue retention rate was as high as about 900%, a figure that would make any software company envious. But when calculating this core metric, individual subscribers to Claude Pro and Max are excluded, on the grounds that fixed-price subscriptions cannot reflect the expansion and contraction behavior that this metric is meant to measure.

What a company cares about most can be seen in what metrics it reports to investors. When a company does not show daily or monthly active users and even excludes individual subscribers when calculating its core retention rate, the weight of ordinary individual users in its mind is self-evident.

02 The money is made by Claude Code and enterprises

Since the bulk is usage-based revenue, what exactly are enterprise customers burning money on? The prospectus shines its spotlight on Claude Code, calling it the product that "most concretely defines what autonomous AI work looks like."

An ordinary chat box doesn't consume much Token. What can really drink computing power like water are agents that can write code, run tests, and fix bugs on their own. The prospectus offers a comparison: in the third quarter of 2026, Claude Enterprise users who did not use Claude Code spent less than $75 per month on average, while the top 5% of Claude Enterprise users who used Claude Code spent about 20 times that of the former.

The prospectus divides product forms into three generations: chatbots, Copilots embedded in a single application, and "virtual collaborators" that can take on goals and work autonomously for long periods | Image source: Anthropic

This is also why Anthropic can leverage tens of billions in revenue with a lean team. As of September 30, 2026, the company had about 5,300 employees, and revenue per employee over the past 12 months exceeded $6.6 million, compared with $1.4 million a year earlier.

Once you understand the operating logic of this money-making machine, look at how it treats free users.

In financial accounting, the computing power consumed by free users is classified under "sales and marketing" expenses. The notes to the financial statements are very clear that this expense includes channel partner fees, marketing and advertising, personnel costs, and "computing power and infrastructure costs provided for free access." In the first half of 2026, sales and marketing expenses were $3.808 billion. In other words, free users are a customer acquisition cost on the books.

The risk factors section puts it even more plainly. The company acknowledges that providing free access is largely to build awareness and usage habits, to reach individuals doing knowledge work outside corporate organizations, and hopes to gradually convert them into paying customers; but the conversion may not succeed, and the free tier may also cannibalize paid subscriptions.

The prospectus also emphasizes twice that the company "has been operating in a computing-power-constrained environment, with customer demand exceeding available computing power." The prospectus does not say how computing power is allocated among different customers, but under the premise that computing power is always insufficient, with fixed monthly individual subscribers on one side and Claude Code enterprise users who spend 20 times more on the other, this calculation is not hard to make.

The "revenue per $1 of computing power" chart in the prospectus: from $0.40 in the first quarter of 2025 to $2.08 in the second quarter of 2026 | Image source: Anthropic

Pro and Max users around the world are all experiencing changes in quotas and rate limits. In response, Anthropic admits in its risk factors: we have changed, and may in the future change, usage limits, rate limits, or model and product capabilities, which may harm our reputation and the value of our products in users' minds; some of these changes have already triggered, and may in the future trigger, allegations that our disclosures about model access or capabilities are inadequate or misleading.

Knowing that changing quotas will draw criticism and damage its reputation, Anthropic still wrote this into the prospectus, giving investors a heads-up in advance.

03 Users as "kindling"

Since individual users contribute so little revenue and also consume constrained computing power, why doesn't Anthropic simply shut down the consumer side and become purely a backend API supplier?

This is exactly the subtle point in the prospectus. Individual users carry little weight on the revenue ledger, but they have irreplaceable strategic value in the chain of business expansion.

They are the kindling that helps Anthropic break into the interior of enterprises.

The prospectus reviews this growth flywheel. So far, the diffusion of Claude has largely been spontaneous, driven by software developers and professional individual users. These people first use Claude themselves, then push for broader adoption within their organizations, saving the trouble of building a huge direct sales team. The company says outright that this approach allows them to achieve scale with a smaller team than other large enterprise technology companies.

Software developers are the most core group among them. The prospectus argues that they understand how systems work, have credibility in evaluating technical tools, and often drive the adoption of new technologies internally; winning their trust is the foundation for Claude's broader diffusion among enterprise employees.

Once this logic is clear, it becomes understandable why free computing power is recorded under "sales and marketing." Providing services to individual users is essentially Anthropic's customer acquisition method: a programmer first uses it themselves, then goes back to the company to persuade the team to purchase the Enterprise plan, and finally connects the entire R&D process to usage-based Claude Code.

But this model has a soft spot. The prospectus acknowledges that it assumes individual adoption will continue to translate into organization-level adoption. In the risk factors, Anthropic warns investors that companies with stronger consumer businesses, larger user bases, and more mature brands may increase their investment in enterprise customers; if these competitors take away "the starting point of our enterprise adoption"—that is, individual users—through consumer apps or other means, the company's business may be adversely affected.

This creates an awkward situation. Anthropic needs individual users to help it build word of mouth and act as promoters inside enterprises, and it fears competitors will take this group away; at the same time, the prospectus clearly states that the company's research priorities, product roadmap, and marketing team are all organized around enterprise use cases.

It needs you to be the kindling, but its core resources all serve enterprises. When large enterprise orders pour in and computing power is stretched, the first to have its oxygen supply restricted is precisely the kindling itself.

04 A To B company dressed in consumer-grade clothing

Close this prospectus, and Anthropic's true face is already clear: it is essentially a To B company. To acquire customers and build an ecosystem, it has made the product a chat box that anyone can pick up, but the real business is selling usage-based computing power and autonomous working Agents to tens of thousands of enterprises.

The product structure chart in the prospectus: at the bottom are the four model series Mythos, Opus, Sonnet, and Haiku; in the middle is the developer platform; at the top is the value delivered to enterprise customers | Image source: Anthropic

As for the account suspension issue that domestic users are most sensitive to, the prospectus also gives an institutional explanation: the company has a "supported regions policy" and implements IP-based geographic blocking to restrict access to its models, platform, and applications in specific regions. Mainland China was never within its service area to begin with.

When a company with a valuation heading toward $2 trillion puts all its chips on enterprise workflows and code agents, the position of individual users in its eyes will only move further and further back.

It is just unclear whether, as more and more developers who have been turned away turn to other competitors, Anthropic can still hold on to the starting point of enterprise adoption that it prides itself on.

This article comes from the WeChat public account "Geek Park" (ID: geekpark), author: Wildcard, editor: Jingyu, authorized for publication by 36Kr.