On Thursday, September 3, artificial intelligence service Grok was down for more than three and a half hours due to an issue at the company’s data center in Memphis, which also caused disruptions to partner platforms.
On Thursday, September 3, artificial intelligence service Grok was down for more than three and a half hours due to an issue at the company’s data center in Memphis, which also caused disruptions to partner platforms.
This is reported by Engadget with reference to a statement from SpaceXAI. The failure, which the site labeled as a «model failure,» began at 6:30 a.m. Pacific Time and affected the work of Grok on the X platform (Twitter), as well as applications for iOS and Android.
«We apologize for any issues you may have experienced while using Grok following an outage at our Memphis data center this morning. We would also like to apologize to our data center partners who were impacted by this situation,» the company said in a statement.
The head of the company, Elon Musk, clarified in his own post that specialists are already taking corrective measures to avoid similar incidents in the future. However, the causes of the problem in Memphis have not yet been disclosed.
In parallel with the problems in Grok , other AI developers also reported failures. Anthropic reported an increased error rate in the Claude Mythos 5.1, Claude Fable 5.1, and Claude Opus 5 models starting at 6:30 a.m. The problem was fixed in two and a half hours. The developers did not confirm a direct connection to the events in Memphis, but Anthropic had previously signed an agreement to lease computing power from Musk’s structures.
Additionally, an increased number of errors in ChatGPT and Codex were reported by OpenAI. According to Downdetector, users began reporting issues around 7:30 a.m. Pacific Time, and systems were fully restored by 9:55 a.m.
The Memphis supercomputer center is a key site for deploying and training artificial intelligence models for Elon Musk’s companies. Recently, the complex’s computing resources have been actively provided for use by third-party technology companies to optimize infrastructure costs.
