IT Home October 27 news: Anthropic is building a cross-product ecosystem — the kind of ecosystem that gives tech companies like Amazon and Uber extremely strong user stickiness — and its S-1 prospectus shows this strategy is already showing early success. Currently, among customers spending at least one hundred thousand US dollars per year, more than half use both the Claude platform and first-party applications such as Claude Code and Cowork; at the end of last year, that figure was only about thirty percent.
The S-1 filing, submitted to the US Securities and Exchange Commission this past Monday local time, states: "The speed and depth of enterprise customer adoption provide compelling proof that Claude is at the frontier of artificial intelligence."
As of September 30, Anthropic had more than 11,900 customers spending no less than 100,000 US dollars per year (IT Home note: about 668,000 RMB at current exchange rates); more than 1,900 customers spending 1 million US dollars (about 6,684,000 RMB at current exchange rates) or more annually; and roughly 250 customers spending 10 million US dollars (about 66,774,000 RMB at current exchange rates) or more per year.
Anthropic attributes its continued growth in customer spending to "customers deploying Claude widely across their enterprises, with its value recognized and usage increasing significantly," which drove the company to achieve a net revenue retention rate of about 900% at the end of the third quarter.
Anthropic tacitly questions the traditional SaaS industry, positioning Claude not as a "static product serving a single scenario" but as a "frontier AI system whose capabilities iterate rapidly," spanning the model layer, developer platforms, and the various applications users use directly.
The filing states: "Technical progress achieved at any layer creates a compounding effect across the three-layer architecture, accelerating the pace at which we deliver stronger intelligence capabilities." The company says it is precisely this integrated approach that is leading more and more customers to adopt both the Claude platform and first-party applications.
This playbook is not unfamiliar. Uber initially offered only ride-hailing before expanding into food delivery and freight; Amazon used its AWS cloud services to attract customers, then leveraged those relationships to extend into storage, databases, and other products, gradually building an ecosystem that customers find hard to leave. Like these companies, Anthropic does not have to find brand-new buyers for every product: an enterprise can start by calling the API, then deploy Claude Code for its engineers, and then roll out Cowork to its entire workforce.
Users of such applications are far more valuable. The prospectus shows that in the third quarter, Claude enterprise customers using Claude Code whose spending ranked in the top 5% paid roughly twenty times more per month than ordinary enterprise customers that did not use this coding tool. The data indicates there is still considerable room for growth in per-customer spending, and explains why Anthropic is so eager to get its first-party applications into the accounts of customers that started with API services.
Anthropic says that with its Applied AI team, it has further amplified this natural penetration effect. The internal Applied AI team works with startups and large enterprises, helping customers embed Claude into their existing technology stacks. Recently, the company also launched a new program called Ode with Anthropic together with Blackstone and other private equity firms, dispatching dedicated engineers into enterprises to build customized AI tools for customers based on Claude.
Anthropic also relies on third-party distribution partnerships with hyperscale cloud providers AWS, Google Cloud, and Microsoft Azure. The filing calls this model a "co-marketing arrangement": Anthropic signs contracts directly with customers, then pays service fees to each cloud provider. The company says this approach can "avoid channel conflicts" and "retain independent control over pricing."
As this AI giant continues to launch "entirely new categories of products, aiming for the new products to create economic value far exceeding the existing business," this still-forming stickiness strategy may prove crucial. The company says that, rather than spending much effort consolidating existing products and business models like ordinary enterprises, it will devote more resources to defining new businesses.
Despite eye-catching revenue growth — revenue of 16.3 billion US dollars (about 108.842 billion RMB at current exchange rates) in the first half of 2026, far exceeding the 1.2 billion US dollars (about 8.013 billion RMB at current exchange rates) in the same period last year — the prospectus also lists a number of headwinds. Cloud partners that help sell Claude are increasingly viewing Anthropic as a competitor, and Anthropic believes this could make it difficult to obtain the computing power it needs from these cloud providers. The filing states that Amazon, Google, and Microsoft may prioritize their own needs, or give priority to other rival products they have invested in.
Customer concentration is another risk. The risk factors section states: "Any adverse change in the demand of top customers ... would negatively affect our business and growth prospects. Accordingly, if one or more key customers substantially reduced their spending or terminated their cooperation, with no new customers or other existing customers increasing orders to fill the gap, it would adversely affect our business, financial condition, and operating results."