Back
RCreddit.com
WARM

ChatGPT, Claude, and Grok all went down within hours of each other yesterday; here's what actually happened

View original
ClaudeOpenAIGrok
Time & source
Published
09/04, 08:03
First discovered
09/04, 15:00
Type
Dev community · RSS
AI summary

ChatGPT, Claude, and Grok experienced outages yesterday, with OpenAI's ChatGPT and Codex going down first around 7:43 AM PT due to a routing error affecting web, mobile, and desktop users. This issue was resolved within approximately half an hour. While there were concerns about an "AI apocalypse," all services were restored by early afternoon PT, and no data loss was reported by any of the three platforms, despite a few hours of disruption.

Saw a lot of panic posts about this yesterday, so figured I'd write up what's actually confirmed instead of the AI apocalypse takes.

OpenAI went first, around 7:43 AM PT; a routing error took out ChatGPT and Codex for a bunch of users across web, mobile, and desktop. They had it fixed in about half an hour.

Claude went down separately and for longer. Anthropic's status page showed elevated errors on Claude.ai, Claude Code, and the API, mostly affecting login and chat completions. That one ran for around three hours before it was fully resolved.

Grok had its own outage in roughly the same window too, confirmed on xAI's status page.

The thing that got everyone tweeting conspiracy theories is the timing overlap, but from what's been reported these are three unrelated incidents on three completely different infrastructures. OpenAI blamed a routing error; Anthropic called it an infrastructure issue. No indication any of them share a root cause or provider.

What's actually striking to me is how many people said some version of How am I supposed to work right now? instead of just being mildly annoyed. These tools clearly aren't side toys anymore for a lot of people; they're load-bearing parts of daily workflows now, which also means single-vendor dependency for anything you're building is a real risk, not a hypothetical one.

Everything was back to normal by early afternoon PT; no data loss was reported by any of the three. Still a wild few hours though.