Frontier models from Anthropic, OpenAI, and xAI all experienced rare outages on Thursday morning. The incident created downtime for their corresponding AI chatbots.
SpaceX said on Thursday afternoon that the issues with Grok resulted from an outage at the Memphis compute center on Thursday morning. The issues initially appeared linked because they coincided. Observers suspected a shared third-party service provider.
Neither OpenAI nor Anthropic cited an external source in comments to WIRED on Thursday. SpaceX did not respond to WIRED for comment. The company apologized to impacted compute partners in public comments on Thursday.
Anthropic and xAI announced a compute partnership with SpaceX in May. Typically, multiple outages in the same sector at the same time point to a cloud provider, content delivery network, or third-party vendor affecting multiple customers.
OpenAI and Anthropic did not point to a potential shared cause. Major internet infrastructure players including Cloudflare, Amazon Web Services, and Microsoft Azure did not report outages on Thursday.
OpenAI spokesperson Kathleen Chaykowski stated that a routing error starting around 7:43 am PT on Thursday, September 3, made ChatGPT and Codex unavailable for some users across platforms. A solution was successfully implemented and monitored as of about 8:17 am PT.
Anthropic declined to comment on the episode. The company began alerting about a partial outage at 6:23 am PT involving elevated errors on requests to Claude Mythos 5.1, Claude Fable 5.1, and Claude Opus 5. The company identified the cause and deployed a fix, marking the issue resolved by 9:16 am PT. Claude Sonnet 5 briefly had similar issues shortly after 9 am PT.
XAI reported Grok outages across all platforms and services beginning at 6:30 am PT when posting investigating outage on its status page. The company stated that traffic was healthy again after resolving the situation at 10:05 am PT.
Scattered reports of a possible Google Gemini outage also occurred on Thursday morning. Google did not confirm this or record incidents on its dashboard, and did not respond to WIRED requests for comment ahead of publication.
OpenAI stated that it is continuing to monitor the solution implemented for its routing error.
