OpenAI, Anthropic, and xAI Hit by Simultaneous Outages

OpenAI, Anthropic, and xAI all suffered rare outages on Thursday morning, disrupting access to their flagship AI chatbots. The incidents occurred within a tight timeframe but the companies did not cite a common external cause.

OpenAI spokesperson Kathleen Chaykowski attributed its problem to a routing error. "A routing error starting around 7:43 am PT on Thursday, September 3, made ChatGPT and Codex unavailable for some users across platforms," Chaykowski said. The company reported implementing a solution by approximately 8:17 am PT.

Service Disruptions Detailed

Anthropic alerted users to a partial outage beginning at 6:23 am PT. The issue involved raise error rates for several of its Claude model variants. The company identified a cause, deployed a fix, and marked the incident as resolved by 9:16 am PT. Its Claude Sonnet 5 model experienced brief, similar issues shortly after 9 am.

XAI reported Grok outages across all platforms starting at 6:30 am PT. The company's status page indicated an investigation was underway. Service was restored by 10:05 am PT. In a public comment, SpaceX, xAI's parent company, said the Grok issues resulted from "an outage at our Memphis compute center this morning." The statement also included an apology to "our impacted compute partners."

Speculation and Silence on Cause

The simultaneous timing initially suggested a potential shared third-party provider failure. Such coordinated outages often point to issues with a major cloud provider, content delivery network, or other infrastructure vendor. However, neither OpenAI nor Anthropic pointed to an external source in their comments to WIRED.

Major infrastructure firms, including Cloudflare, Amazon Web Services, and Microsoft Azure, reported no widespread outages on Thursday. This absence of reports from key internet backbone companies deepened the mystery surrounding the coincidental timing.

Other Services and Partnership Context

There were scattered, unconfirmed user reports of issues with Google's Gemini AI on Thursday morning. Google did not confirm any outage or log an incident on its official service status dashboard. The company did not respond to requests for comment.

The incident follows a previously announced compute partnership between Anthropic, xAI, and SpaceX. This business relationship adds a layer of complexity to understanding the independent nature of the September outages.

The table below summarizes the outage timelines for the three confirmed incidents.

Anthropic declined to comment further on the episode. SpaceX did not respond to WIRED's request for additional details beyond its public statement.