Anthropic capacity crisis: Claude down five hours, usage restrictions introduced

On Friday, March 27, 2026, Anthropic experienced its longest outage to date. Claude API, Claude Chat, the Claude app, and Claude Code were all unavailable for nearly five hours due to what the company described as "networking performance degradation within cloud infrastructure." Communication between components in the serving stack broke down. Normal service was restored at 9:30 AM PT after affected workloads were migrated to healthy infrastructure. It was the second outage in a single week.

The disruption came just days after Anthropic announced a new capacity policy that had already drawn user frustration. Starting March 27, the company began enforcing what it calls "usage limit adjustments during peak hours." In practice, this means Free, Pro, and Max subscribers now exhaust their five-hour session limits more rapidly between 1 PM and 7 PM GMT on weekdays. Weekly limits remain unchanged, but the impact during busy hours is significant. According to Anthropic, roughly 7% of Pro users will experience this directly.

The company had simultaneously run a temporary promotion from March 13 to March 28 doubling limits during off-peak hours, which partially offset the restrictions. But the combination of throttling and a major outage has triggered genuine frustration in the user community. On Hacker News and Reddit, many reported unexpected rapid depletion of Claude Code quotas during the same period, without clear communication from Anthropic.

The situation points to a broader problem: demand for Claude is growing faster than capacity. With GPT-5.4 and Gemini 3.x as direct competitors, the timing is difficult for Anthropic to struggle with uptime. The company has not commented on exactly what triggered the capacity pressure, but industry observers are now watching whether this is an isolated incident or a systemic capacity deficit that will continue affecting the service.

For enterprises and developers who have integrated Claude into production systems, this serves as a reminder of the importance of fallback strategies and redundancy in AI architecture.