ChatGPT, Claude, Grok and Gemini all suffered overlapping outages the same day

Started by Zenith Baz, Yesterday at 01:47 PM

Previous topic - Next topic

0 Members and 1 Guest are viewing this topic.

Topic: ChatGPT, Claude, Grok and Gemini all suffered overlapping outages the same day   Views(Read 42 times)
Active members in this topic:
Zenith Baz(1)

Zenith Baz

Millions of users hit error screens on September 3rd as ChatGPT, Claude, Grok and Gemini all experienced disruptions within a similar window, an overlap across four separately developed AI models from competing companies that's exceptionally rare even though single platform outages are common. Downdetector logged over 14,000 cumulative user reports at the peak, with OpenAI's reports exceeding 12,000, alongside roughly 1,200 for Claude and a smaller but still notable number for Grok, while Google didn't issue a formal outage notice despite user reported disruptions

Anthropic's Claude dashboard showed outages specifically across its Opus 4.8 and Opus 5 models, with other Claude models reported to have recovered to baseline error rates. OpenAI confirmed elevated errors across ChatGPT and its Codex coding tool, while xAI acknowledged Grok was experiencing issues. Secondary services that depend on these underlying models, including the AI coding tool Cursor, were also affected as their upstream infrastructure struggled. Early speculation pointed toward shared dependencies like regional cloud backbone providers, since all three of OpenAI, Anthropic and xAI reportedly rely on Microsoft Azure for compute capacity, though no company had confirmed a root cause

Independent monitoring data shows this kind of disruption is more common than press releases typically suggest, major AI providers maintained availability at or above 99.9 percent across 2026 overall, but that still allows for nearly 9 hours of downtime a year, and providers collectively experienced dozens of similar incidents tied to scaling pressure on inference infrastructure. Recovery times were swift, measured in minutes to a couple of hours across the affected services. Curious what people think this reveals about how much shared infrastructure sits underneath supposedly independent, competing AI companies


Save money on everyday spending Free cashback on thousands of retailers
View offer