Anthropic Admits Claude Code Users Hitting Usage Limits 'Way Faster Than Expected'

Anthropic has acknowledged that Claude Code users are exhausting their quotas far more quickly than anticipated, disrupting developer workflows across its platform.

The Problem

Users across multiple subscription tiers report dramatically reduced effective usage:

• Pro plan ($200/year): One user reports maxing out every Monday, only ~12 usable days per month
• Max 5 plan ($100/month): Another user exhausted their quota in just 1 hour of work, down from 8 hours previously

Contributing Factors

1. Peak-hour throttling: Anthropic reduced quotas during peak hours last week, affecting ~7% of users
2. Promotion expiration: March 28 was the last day of a doubled-usage promotion outside a 6-hour peak window
3. Potential bugs: A user claims to have found two bugs in the Claude Code binary that break prompt caching, "silently inflating costs by 10-20x". Downgrading to version 2.1.34 reportedly helped

Prompt Cache Economics

The prompt cache has a 5-minute lifetime by default:

• Stopping for a short break invalidates the cache
• Upgrading to 1-hour lifetime costs 2x base input token price
• Cache reads are only 0.1x base price, making breaks extremely costly

Analysis

This is a significant growing pain for AI coding assistants. Claude Code's rapid adoption has created demand that Anthropic's infrastructure struggles to meet. The prompt cache bug — if confirmed — represents a serious efficiency problem where developers are being charged for tokens they shouldn't need. The opacity of Anthropic's usage limits (never specifying exact quotas) compounds user frustration.

Anthropic's position as a leader in AI safety ironically creates a trust gap when users can't predict their costs or usage. The company needs transparent per-plan quotas and reliable prompt caching to maintain developer confidence.