Claude's code popularity is forcing Anthropic to limit access
Users will get weekly limits on their Claude Code usage to curb power users and stabilize its infrastructure.
If you've been running Claude Code like it's your full-time engineering assistant, Anthropic is about to slow you down.
Anthropic is officially rolling out weekly rate limits for Claude Code starting August 28 on its Pro and Max subscriptions. And while this might only affect the top 5% of users, according to Anthropic, it’s obvious that Claude Code is in high demand, and resources are tight.
This change is mostly targeting users who’ve been running Claude in the background 24/7, chewing through compute endlessly. In some cases, those users have been sharing or reselling accounts. But the bigger issue is a strain on infrastructure, affecting the need to make sure everyone gets a reliable Claude experience.
We’re rolling out new weekly rate limits for Claude Pro and Max in late August. We estimate they’ll apply to less than 5% of subscribers based on current usage. pic.twitter.com/X8FAss3jIC
— Anthropic (@AnthropicAI) July 28, 2025
Under this new structure, the weekly limits will supplement the existing five-hour rolling limits. The caps vary based on subscription tier. $20/month Pro users get 40–80 hours of Claude Code time weekly with the Sonnet 4 model. The $100/month Max users get 140–280 hours of Sonnet and 15–35 hours of Opus 4, while the $200/month Max users stretch to 240–480 Sonnet hours and 24–40 Opus hours.
Now, the math is quite interesting. Anthropic has previously advertised the $200 Max plan as offering 20x the usage of the Pro tier. But looking at these new hourly estimates, you’re only getting about 6x the Claude Code hours. So, unless that 20x is measured in tokens or raw compute, which Anthropic hasn’t clarified, the numbers aren’t adding up.
Emmanuel Oyedeji
And Claude Code has been under pressure. It’s seen at least seven service outages in the past month alone, according to Anthropic’s status page.
This isn’t a Claude-only problem. Other AI dev tools are hitting similar walls. Cursor recently restructured its Pro plan pricing to reduce abuse. Replit has also reshaped its pricing strategy in response to overwhelming demand from heavy users. The whole space is grappling with a familiar problem of AI tools becoming incredibly powerful, but not infinite.
These limits carry competitive implications. Developers are increasingly central to the AI arms race. Rate limiting, especially for a tool like Claude code, could push serious devs to explore alternatives. If Claude becomes something you can’t reliably use all day, every day, it opens the door for others to catch up.
Ogbonda Chivumnovu
