Anthropic's entire Claude product line went dark on April 28. The outage lasted 78 minutes, from 17:34 to 18:52 UTC. Everything was hit: Claude.ai, the API, and Claude Code, Claude Cowork, Claude Console, and Claude for Government. Authentication errors blocked users from logging in and broke API requests. Services recovered by 18:59 UTC, according to Anthropic's status page.
One enterprise user on Hacker News reports spending $200,000 per month on Anthropic's services. Their uptime over the past 90 days? Only "one 9." That's approximately 90% availability, or roughly 36 hours of downtime per month. Enterprise customers typically expect 99.9% uptime, about 43 minutes of downtime monthly. The gap isn't small. It's enormous. Multiple commenters echo the frustration. A previous widespread outage. Some say their executive teams are losing patience.
The fact that consumer, developer, enterprise, and government products all failed simultaneously points to a shared infrastructure component. A single point of failure across your entire stack is a serious architectural problem, especially when you're serving government agencies that require 99.9% or better uptime. Anthropic is growing fast, and scaling infrastructure is genuinely difficult. But customers paying six figures monthly shouldn't be getting 90% availability. That's a reliability crisis.