Cursor cancelers’ direct Claude rate rose from below to above the rate for retained users
Share with an active direct Anthropic or Claude subscription in March–May 2025 and March–May 2026
Share with a direct Claude subscription
Source: YipitData U.S. e-receipt panel. Cursor cancelers n=324; retained Cursor paid users n=1,125. The canceler-minus-retained change was +11.3 percentage points. Cursor cancellation was identified from a lifecycle email, and Claude’s broader growth was present in both groups.
Chart data
| Series | Share with a direct Claude subscription |
|---|---|
| Cursor cancelers — Mar.–May 2026 | 54.3% |
| Retained Cursor users — Mar.–May 2026 | 49.6% |
| Retained Cursor users — Mar.–May 2025 | 26% |
| Cursor cancelers — Mar.–May 2025 | 19.4% |
How did direct Claude subscriptions change around Cursor cancellation?
In a separate view aligned to each user’s cancellation month, the direct Claude subscription rate among 324 Cursor cancelers was about 15% three months before cancellation, about 30% in the cancellation month, and 39% six months afterward. Rival developer-tool subscriptions fell from about 20% to 12%, while direct OpenAI or ChatGPT subscriptions stayed near 4%. These paths do not establish that Claude functionally replaced Cursor.
How often did Cursor and Anthropic payers also pay for the other service?
From September 2025–February 2026, 37.4% of paying Cursor panelists also paid Anthropic directly, while 5.1% of Anthropic’s paying base also paid Cursor. Direct OpenAI subscriptions appeared for 10.7% of Cursor users. The percentages show how often both subscriptions appeared in observed receipts; they do not measure product dependence or primary usage.
Assumptions & Methodologies
We analyzed 324 established U.S. panelists with a Cursor cancellation email from July 2025–January 2026 and 1,125 paid Cursor users with no observed cancellation. Direct Claude meant a recurring Anthropic or Claude receipt. We compared March–May 2025 and March–May 2026. Cancelers rose 34.9 points versus 23.6 points for retained users, an 11.3-point difference in increases. Cancellation emails are lifecycle signals rather than billing records, and some retained users may have silently lapsed. Partial inbox coverage makes levels floors. GitHub includes non-Copilot billing. Results do not establish that pricing caused cancellation or Claude replaced Cursor usage.