Subscription Decline? Claude Code Growth Plummets from Over 600% to 5.2% — Retention Rate Now Key

Claude Code saw explosive growth in the first half of this year, but external tracking data shows a clear slowdown starting in August. Recent measures by Anthropic—such as introducing invisible watermarks for AI-generated content and adjusting usage quotas for Claude Code—have sparked discontent among some developers, making OpenAI's Codex one of its main competitors. It should be noted that some online claims lack independent verification, so the market should distinguish between 'slowing growth' and 'mass cancellations.'

According to TickerTrends tracking data, Claude Code's tracked Annual Recurring Revenue (ARR) was approximately $2.5 billion on February 12, 2026. It rose to $6.3 billion by the end of March, surpassed $10.3 billion by the end of April, and exceeded $14 billion in June—nearly sixfold growth in just a few months. However, as of August 10, tracked ARR stood at about $15.12 billion. While still substantial—accounting for roughly 21.9% of Anthropic’s total tracked ARR—the recent monthly growth rate has dropped to 5.2%, clearly signaling a cooling of the earlier rapid expansion.

This change does not mean Claude Code has lost market relevance. Rather, it suggests the product may be transitioning from an early rapid penetration phase into a more competitive, mature stage. Claude Code initially gained traction in the developer market through its long-context capabilities, code generation, and agent-style workflows. However, OpenAI has been steadily enhancing Codex over recent months, shifting competition beyond model performance alone to include pricing, usage quotas, IDE integration, and enterprise ecosystems.

TickerTrends previously reported that after a major update to OpenAI Codex at the end of April, download volumes surged. For the week ending May 3, Codex downloads reached 86.1 million—an increase of 1,397% week-on-week—while Claude Code downloads were around 7.2 million, down 38% from the previous week.

In August, Anthropic faced another wave of user backlash. On August 2, 2026, or later, new Claude models began embedding machine-detectable, invisible watermarks into generated text, while images received digital provenance information. This measure aligns with the EU AI Act’s transparency requirements for AI-generated content. Since watermarking occurs at the model level, it affects not only the Claude chatbot but also products using new models such as Claude Code and the API.

Anthropic states that text watermarks do not alter meaning, quality, or readability and are designed to persist through copy-paste and partial editing. Nevertheless, this policy has drawn criticism from the developer community. Some users worry that even when using Claude for proofreading, debugging, or minor edits, their final work might retain identifiable AI-processing traces.

Indeed, some developers have begun researching ways to remove these invisible markers. Wired reported that open-source tools have recently emerged attempting to disrupt watermark patterns through rewriting or reordering text, attracting significant attention. However, the effectiveness of these tools remains unverified, and removing AI provenance markers could pose legal, contractual, and policy risks.

Meanwhile, Claude Code’s usage quotas have become a pain point for heavy users. In August, Anthropic announced that the automatic mode for Claude Code would become the default setting for Pro, Max, and Team plans, and stated it would no longer charge users for additional tokens generated by the automatic mode classifier.

Recently, however, the community has expressed heightened concern over the upcoming end of a 50% extra usage quota benefit. Some Claude Code users on social platforms say that if the bonus quota ends, their effective available usage could drop significantly, prompting them to consider switching to competing products like Codex. These are individual experiences and opinions, not direct evidence of mass cancellations at Anthropic.

Business Insider recently reported that dozens of Claude users have canceled subscriptions due to the invisible watermark issue, with some developers, consultants, and researchers shifting to other AI tools. However, Anthropic says it has not observed a noticeable rise in cancellation rates, so there is currently insufficient evidence to confirm the alleged 'mass cancellation wave' circulating online.

Another challenge for Claude Code is expanding beyond developer tools into broader white-collar workflows. Anthropic has continued to expand Claude Cowork’s functionality, integrating Claude’s Chrome sidebar feature into Claude Cowork on August 12—indicating the company’s effort to extend Claude from a programming tool to general workplace applications.

However, the competition between Claude Code and Codex is no longer just a battle between two products. OpenAI benefits from ChatGPT’s massive user base and enterprise sales infrastructure, while Anthropic continues expanding in enterprise AI and programming domains. For developers, subscription decisions depend not only on the model’s coding ability but also on usage quotas, pricing, IDE integration, stability, enterprise compliance, and data handling policies.

It would be premature to interpret Claude Code’s slowing growth as the end of Anthropic’s AI narrative. A recent Financial Times report notes that investors still bet on Anthropic potentially going public in October 2026, with a valuation possibly exceeding $2 trillion, and anticipate the company’s annualized revenue could reach $100–120 billion by year-end. The report also points out that Anthropic faced regulatory-driven growth slowdowns in June but subsequently re-accelerated.

Therefore, what matters most for Claude Code now is not whether its 'myth' has been shattered, but whether Anthropic can maintain high retention rates in a more fiercely competitive market after the hypergrowth phase ends. As OpenAI pushes Codex forward, rival AI programming tools vie for developers, and AI watermarking and usage restrictions spark new product experience debates, Anthropic must prove that Claude Code’s technical advantages outweigh friction from pricing, quotas, and policy constraints.

From rapid ARR growth at the start of the year to single-digit monthly growth in August, Claude Code is indeed facing a different market environment than in its early days. But for a company still in a high-growth phase, a 5.2% monthly growth rate alone does not indicate product decline, nor can it directly imply widespread user exodus. The truly critical metrics will be ARR growth over the coming quarters, paid user retention, enterprise customer expansion, and whether competing products like Codex can continue taking developers from Anthropic.