On Tuesday, April 15, Claude went dark for roughly three hours. No login, no API, no Claude Code. Over 5,000 users reported problems simultaneously on Downdetector.
The timeline
At 10:53 AM ET, Anthropic first acknowledged the issue: ‘Elevated errors on Claude.ai, API, Claude Code.’ Ten minutes later, they announced a fix — which didn’t hold. By 11:40 AM, the status page went red: ‘Claude.ai and Platform are down.’
Everything was back online at 1:42 PM ET. Three hours total, right in the middle of the American workday.
What was affected
Pretty much everything. Claude.ai was unreachable, the API returned 500 errors, login verification codes failed to arrive, and prompt limits showed incorrect values. Both free and Pro users were hit.
One silver lining: users already logged into Claude Code could keep working. New logins, however, were blocked.
The pattern
This wasn’t an isolated incident. Similar issues hit in March and early April — often following major model updates or demand spikes. Anthropic’s infrastructure seems to be struggling to keep pace with its rapid user growth.
My take
A three-hour outage is annoying but not catastrophic. What concerns me more is the pattern. If you’re betting your workflow on Claude — and more companies are doing exactly that — you need reliability. Anthropic recently announced a partnership with CoreWeave to expand compute capacity. That expansion can’t come fast enough.
Sources: