SAN FRANCISCO, Sept. 7 — A cascade of outages swept through several major AI platforms in a single morning, briefly knocking out service for users of OpenAI’s ChatGPT, Anthropic’s Claude, Google’s Gemini, and X’s Grok. The trouble surfaced in waves. Downdetector reports showed OpenAI services starting to falter around 10:30 a.m. ET, with complaints peaking above 36,000.
OpenAI’s status page said a fix had been applied and recovery was being monitored after about 24 minutes of downtime across ChatGPT and its Codex tool. By 1:25 p.m. ET, the company declared full operational status, saying it was “not aware of any issues affecting our systems,” and reports on Downdetector had tapered off.
Gemini also appeared to be running normally again by then. Claude’s problems began earlier, with a small spike around 8:30 a.m. ET that surged an hour later. By 3:15 p.m. ET, Anthropic’s status page showed “All systems operational.” Grok was the straggler.
Hours after its rivals recovered, users were still hitting “Grok was unable to reply” errors, and the chatbot displayed a warning that it was “experiencing issues” and that “some features may be slow or unavailable,” directing people to status.x.ai — which initially showed no disruptions. At 4:00 p.m. ET, SpaceXAI apologized to users and compute partners, blaming “an outage at our Memphis compute center this morning.” The company said all systems had been restored and were “functioning nominally,” and queries on Grok subsequently went through.
The companies did not immediately disclose what caused the broader outages. The episode underscored how tightly coupled modern AI services are to shared data-center infrastructure — when one piece of that chain fails, several platforms can feel it at once. For ordinary users, the practical upshot is simple: these tools are only as reliable as the physical data centers behind them.
The question now is whether providers will harden that infrastructure — or whether mornings like this one become a recurring feature of the AI era.
