For two weeks in March, Claude subscribers got double their normal usage limits during off-peak hours. Anthropic called it "a thank you" to users. Then the promotion ended on March 28, and the complaints started rolling in.
The timing was brutal. Right as the 2x bonus disappeared, several other things happened at once: Anthropic quietly reduced peak-hour quotas on March 26, Claude Code turned out to have bugs that inflated token usage by 10-20x through prompt cache failures, and paid subscriptions had more than doubled in early 2026 thanks to features like Claude Code and Computer Use. More users, same capacity, no more bonus.
Max 5x subscribers - paying roughly $100/month - reported burning through their weekly allocation in a fraction of the time they used to. The promotion had effectively reset expectations: what felt generous during the bonus period now felt restrictive by comparison.
Anthropic's Product Lead Lydia Hallie posted an acknowledgment that "people are hitting usage limits in Claude Code way faster than expected," but the advice boiled down to switching to lighter models and reducing effort levels. Users took that as being told to use less of the product they were paying premium prices for.
The original promotion ran March 13-28, covered Free, Pro, Max, and Team plans (not Enterprise), and applied automatically during off-peak hours outside 8 AM-2 PM ET on weekdays. There was also a separate $50 free usage credit offer for users who enabled "Extra Usage" by mid-February.
This is a growing pattern with AI subscriptions: companies offer generous limits to attract users, then struggle to maintain those limits as adoption grows. Anthropic's Claude is clearly gaining traction fast, but the gap between what power users expect and what the infrastructure can deliver is widening. For anyone budgeting around Claude for daily work, the real cost isn't just the subscription price - it's the unpredictability of how much you can actually use.