After reading the prospectus, I finally understand why Anthropic doesn't care about users anymore.
The big one is finally here.
On October 26, Anthropic publicly filed its S-1 registration statement with the U.S. Securities and Exchange Commission, preparing to list on the Nasdaq Global Select Market under the ticker symbol ANT. According to media reports, the company's valuation discussion range has reached $1.5 trillion to $2 trillion, with the offering price range yet to be disclosed.
Within 24 hours, this 300-plus-page prospectus was thoroughly dissected by both the tech and financial sectors. Everyone has seen the most eye-catching figures: revenue reached $16.286 billion in the first half of 2026, compared to just $1.163 billion in the same period last year; the second quarter posted $974 million in operating income, marking the company’s first historically profitable operating quarter; as for the staggering $80.7 billion net loss on the books, the bulk stems from the $72.388 billion fair value adjustments on convertible bonds, derivative liabilities, and SAFEs, which are non-cash accounting charges.
Beyond these Wall Street narratives, domestic developers and everyday users experience a different sense of disconnection. In the eyes of mainland China users, Claude is likely currently the most account-banned, strictly risk-controlled, and frustrating-to-use AI product on the market.
Why does a company that builds top-tier large language models always seem so cold toward ordinary individual users?
After reviewing the financial notes and risk factors in the prospectus, the answer is actually written right in the ledger. In Anthropic’s business landscape, the actual cash contribution from individual users is very small.
01 Under 20% Individual Users
To understand who a company truly cares about, the most direct method is to examine its revenue composition table. In footnote 2 of the prospectus's financial statements, Anthropic splits its revenue into two major categories: usage-based billing revenue and subscription revenue.
The data comparison is stark. For the full year of 2025, usage-based billing revenue was $3.796 billion, while subscription revenue was only $789 million, a ratio of approximately 83% to 17%. By the first half of 2026, usage-based billing revenue surged to $13.064 billion, with subscription revenue at $3.222 billion, meaning the former still accounted for roughly 80%, leaving subscription revenue at only about 20%.
A more critical detail lies within the composition of the subscription revenue. This 20% includes both individual users' Claude Pro and Claude Max subscriptions, as well as enterprise-level Team and Enterprise plans, but 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 possibly much less.
So where does the 80% usage-based billing revenue come from? The answer is cloud provider channels and tens of thousands of enterprises.
Revenue sold to customers through marketplaces of cloud providers like Amazon, Google, and Microsoft has soared: approximately 11% in 2023, 32% in 2024, 47% in 2025, and about 48% in the first half of 2026. Nearly half of the entire picture is supported by distribution pipelines from these three cloud giants.
Supporting this money machine is a rapidly expanding group of institutional clients. As of September 30, 2026, the number of clients contributing over $100,000 in revenue over the past 12 months (counted by independent organizations) skyrocketed from over 950 a year ago to over 11,900. The client structure is also healthy: for the first nine months of 2026, no single client accounted for more than 10% of revenue, and the top five clients combined represented about 25%.
Enterprise customer growth chart in the prospectus: Clients contributing over $100,000 in revenue grew from approximately 20 at the end of 2023 to over 11,900 by the end of September 2026 | Image source: Anthropic
The prospectus uses a specific term for individual paid users: 'prosumers', defined as individual users who subscribe to AI tools for knowledge work, such as researchers, developers, analysts, and writers. The title sounds professional, but this group remains on the fringes when faced with core financial metrics.
In the entire 300-plus-page prospectus, there is no disclosure of daily or monthly active user counts, nor is the number of paid subscribers mentioned.
The metric that best reflects this attitude is net revenue retention.
As of September 30, 2026, Anthropic’s net revenue retention rate reached approximately 900%, a figure that makes every software company envious. However, when calculating this core metric, individual subscribers for Claude Pro and Max were excluded, on the grounds that fixed-price subscriptions do not reflect the expansion and contraction behaviors the metric aims to measure.
What a company cares about most is reflected in the metrics it reports to investors. When a company fails to showcase daily and monthly active users, and even excludes individual subscribers from calculating its core retention rate, the weight of ordinary individual users in its mind is self-evident.
02 The Money Comes from Claude Code and Enterprises
Since the bulk of revenue comes from usage-based billing, what are enterprise customers burning money on? The prospectus focuses the spotlight on Claude Code, calling it the product that "most specifically defines what autonomous AI work looks like."
A standard chat interface doesn't consume many tokens. What truly guzzles computing power like water are agents capable of writing code, running tests, and fixing bugs themselves. The prospectus provides a comparison: in Q3 2026, Claude Enterprise users not using Claude Code spent an average of less than $75 per month, whereas the top 5% of Claude Enterprise users employing Claude Code spent roughly 20 times that amount.
Prospectus categorizes product forms into three generations: chatbots, Copilots embedded in single applications, and 'virtual collaborators' that can take on objectives and work autonomously for extended periods | Image source: Anthropic
This is also why Anthropic can leverage a lean team to drive billions in revenue. As of September 30, 2026, the company had approximately 5,300 employees, generating over $6.6 million in revenue per person over the past 12 months, up from $1.4 million a year ago.
Understanding the operating logic of this money-making machine reveals how it treats free users.
In financial accounting, the computing power consumed by free users is categorized under 'Sales and Marketing' expenses. The financial notes state clearly that this expense includes channel partner fees, marketing and advertising, personnel costs, and "compute and infrastructure costs provided for free access." In the first half of 2026, sales and marketing expenses totaled $3.808 billion. In other words, free users are recorded as a customer acquisition cost on the books.
The risk factors section speaks even more plainly. The company admits that providing free access is largely intended to build awareness and usage habits, reaching individuals doing knowledge work outside corporate organizations, with hopes of gradually converting them into paying customers; however, conversion may fail, and the free tier might even cannibalize paid subscriptions.
The prospectus also emphasizes twice that the company "operates in a compute-constrained environment where demand exceeds available supply." It does not specify how compute is allocated among different customers, but given that compute will forever remain insufficient, balancing fixed-monthly individual subscribers against enterprise users spending 20 times more on Claude Code makes the arithmetic quite clear.
'Revenue generated per $1 of compute' chart in the prospectus: rising from $0.40 in Q1 2025 to $2.08 in Q2 2026 | Image source: Anthropic
Users worldwide on Pro and Max plans are experiencing changes to quotas and rate limits. Addressing this, Anthropic candidly states in its risk factors: We have changed and may continue to change volume caps, rate limits, or model and product capabilities, which could harm our reputation and the perceived value of our products among users; some of these changes have already sparked, and may in the future spark, allegations that our disclosures regarding model access or capabilities are inadequate or misleading.
Aware that changing quotas invites backlash and damages reputation, Anthropic still included this clause in the prospectus, giving investors an advance warning.
03 Users as 'Fire Starters'
If individual users contribute so little revenue and still consume constrained compute, why doesn't Anthropic simply shut down the consumer side and become entirely a backend API supplier?
This is precisely the nuance in the prospectus. Individual users carry very little weight on the revenue ledger, yet hold irreplaceable strategic value in the chain of business expansion.
They serve as the 'fire starters' helping Anthropic breach into internal corporate environments.
The prospectus outlines this growth flywheel. To date, Claude's diffusion has been largely organic, driven by software developers and specialized individual users. These individuals adopt Claude themselves first, then push for broader adoption within their respective organizations, saving the company the hassle of building a massive direct sales force. The company explicitly states that this approach allows them to scale with a smaller team than other large enterprise technology companies.
Software developers form the most critical group within this dynamic. The prospectus argues that they understand how systems operate, hold credibility in evaluating technological tools, and frequently champion new technologies internally; winning their trust is the foundation for Claude's broader diffusion among corporate employees.
With this logic clarified, it becomes easy to understand why free compute is bookkeeping under 'Sales and Marketing.' Providing services to individual users is fundamentally Anthropic's customer acquisition strategy: a developer starts using it personally, returns to the company to persuade the team to purchase Enterprise plans, and ultimately integrates the entire development workflow into usage-based billed Claude Code.
But this model has a vulnerability. The prospectus acknowledges that it assumes individual adoption will continue to translate into organizational adoption. Within the risk factors, Anthropic warns investors that companies with stronger consumer businesses, larger user bases, and more mature brands may increase investments targeting enterprise clients; if these competitors steal away the 'on-ramp to our enterprise adoption,' namely individual users, through consumer apps and other means, the company's business could be adversely affected.
This creates a somewhat contradictory situation. Anthropic needs individual users to build word-of-mouth and act as internal champions in enterprises, fearing competitors will poach this group; meanwhile, the prospectus explicitly states that the company's research priorities, product roadmap, and marketing teams are all organized around enterprise use cases.
It wants you to be the fire starter, but its core resources are dedicated to serving enterprises. When large enterprise deals flood in and compute remains stretched thin, ironically, those receiving the strictest oxygen restrictions will be the fire starters themselves.
04 A To-B Company Clad in Consumer Packaging
Closing this prospectus reveals Anthropic's true nature quite clearly: it is essentially a To-B company. For customer acquisition and ecosystem building, it has shaped its product into an accessible chat interface, but its real business lies in selling usage-based compute and self-operating Agents to tens of thousands of enterprises.
Product structure diagram in the prospectus: the bottom layer consists of four model families—Mythos, Opus, Sonnet, Haiku; the middle layer is the developer platform; the top layer represents the value delivered to enterprise clients | Image source: Anthropic
As for the account banning issue that concerns domestic users most sensitively, the prospectus also offers a procedural explanation: the company has 'supported regions policies' and implements geo-blocking based on IP addresses to restrict specific regions from accessing its models, platforms, and applications. Mainland China was never within its service region scope to begin with.
When a company valuing nearly two trillion dollars places all its bets on enterprise workflows and coding agents, the position of individual users in its eyes can only recede further and further.
Only time will tell whether Anthropic can still guard its proud on-ramp to enterprise adoption as more and more rejected developers turn to competing alternatives.