Anthropic is building an Amazon-like moat around its enterprise business, according to disclosures in its initial public offering filing. More than half of the AI company's six-figure customers now use both its Claude API platform and first-party applications, a cross-selling strategy that locks in revenue and reduces churn.
The strategy works by making Anthropic harder to leave. When a customer commits to Claude for API integration and then adopts one of Anthropic's vertical applications, switching costs compound. They lose not just the API connection but also workflow continuity and application-specific data. This is the same playbook Jeff Bezos perfected at Amazon, where AWS customers who use first-party services like DynamoDB or SageMaker rarely migrate.
Anthropic's S-1 filing, which the company submitted ahead of its IPO, provides rare transparency into its enterprise unit economics. The company has crossed a threshold where its largest customers generate meaningful value concentration. This customer tier, defined as those spending over $100,000 annually, represents the revenue backbone Anthropic needs to demonstrate sustainable profitability to public markets.
The cross-selling approach matters because the AI infrastructure market remains brutally competitive. OpenAI dominates with ChatGPT and its enterprise API. Google Claude competes with Gemini. Xai has Grok. In this landscape, Anthropic cannot compete purely on model quality or API pricing. Instead, it's bundling Claude the foundation model with applications that create stickiness, forcing customers to think in terms of switching an entire suite rather than replacing a single API call.
Anthropic's first-party apps likely include Claude in Slack, which launched as a native integration, and potentially domain-specific vertical tools. By controlling both the platform layer and application layer, Anthropic captures more of the customer's budget and more data about how customers actually use Claude. That data feeds back into model training and product development.
This strategy also de-risks the core API business. If customers use Claude primarily for inference via an API, pricing pressure from competitors could compress margins quickly. But if those same customers also license an application, revenue becomes less price-elastic. They're paying for function and convenience, not just compute.
The IPO filing raises the question of what comes next. Anthropic likely plans to expand its vertical application portfolio post-public. The company may target specific industries where it has high customer concentration: finance, legal, healthcare, customer service. Each vertical represents an opportunity to build a Claude-native application that becomes indispensable to a customer's workflow.
Enterprise customers spend months or years integrating APIs. Switching means rewriting code, retraining teams, and rebuilding workflows. Anthropic's dual-platform approach exploits that friction. By the time a customer has Claude embedded in their operations through both API and application, the cost of defection feels too high. That's what stickiness means in infrastructure, and that's what Anthropic is selling to public market investors.