ChatGPT experienced a significant outage on Thursday afternoon, leaving many users unable to access the AI ​​chatbot or its related services. Around 1:30 p.m. ET, the reports began pouring in, with users encountering reports of “internal server errors.” The disruption also impacted OpenAI’s API and text-to-video generator, Sora, highlighting these tools’ reliance on seamless backend operations.

OpenAI confirmed the issue on its status page at 2:00 p.m. ET, stating, “ChatGPT, the API, and Sora are currently experiencing high error rates.” At 6:15 p.m. ET, the company confirmed that Sora was “fully operational,” while the APIs began to recover. However, the chatbot itself was still having problems, and engineers continued to work on a solution.

Although OpenAI did not name the specific “upstream provider” associated with the outage, Microsoft, its exclusive cloud partner, reported a power outage at its data center in the southern United States. The issue, which began around the same time as ChatGPT’s downtime, caused memory latency, timeouts, and HTTP 500 errors. Microsoft’s cloud gaming service Xbox Cloud was also affected.

At 5:00 p.m. ET, Microsoft reported that power was “fully restored” to the affected data center, but the incident highlights the cascading impact of infrastructure failures on dependent services.

This is not the first time in recent months that ChatGPT has gone offline. In early December, ChatGPT and Sora experienced extended downtime shortly after the video generation tool was launched for subscribers. In June, there was a widespread disruption to several AI tools, including ChatGPT, highlighting the challenges of scaling services that rely on complex cloud infrastructure.