On Thursday afternoon, many users found ChatGPT unresponsive as the chatbot abruptly stopped processing queries. Some users encountered an “internal server error” message, signaling a significant outage.
Reports of the issue began to surface around 1:30 PM ET, with complaints spiking on the outage-tracking website Down Detector. By 2 PM ET, OpenAI acknowledged the disruption on its status page, noting that ChatGPT, its API, and the text-to-video generator Sora were “currently experiencing high error rates.” OpenAI later reported that Sora was restored at 6:15 PM ET, while ChatGPT achieved a “full recovery” by approximately 11 PM ET.
Although OpenAI did not disclose specific details about the “upstream provider” linked to the problem, its exclusive cloud partner, Microsoft, reported a concurrent “power issue” at one of its data centers. The issue, which impacted services in North America, coincided with outages affecting Xbox cloud gaming.
Details on the Power Incident
Microsoft provided an update on the situation, revealing that the problem originated in the South Central US region, specifically in availability zone AZ03. The incident began at 18:44 UTC (1:44 PM ET) on December 26, 2024, and affected multiple services. Impacted users experienced storage latency, timeouts, or HTTP 500 errors.
Microsoft’s statement read:
Impact Statement: “Starting at 18:44 UTC on 26 Dec 2024, you have been identified as a customer who was impacted by a power incident in South Central US and may experience a degraded experience.”
Current Status: “There was a power incident in the South Central US AZ03 which affected multiple services. We have applied mitigation and are actively validating recovery to the impacted services. Further updates will be provided in 60 minutes, or sooner as events warrant.”
At 5 PM ET, Microsoft announced that power to the affected data center had been fully restored.
Recent History of Outages
This incident is not an isolated event for ChatGPT, as the service has experienced several outages in recent months. Earlier in December, shortly after OpenAI introduced the Sora video generation tool to ChatGPT subscribers, both the chatbot and Sora experienced significant downtime. Additionally, a major outage in June disrupted the functionality of various AI tools, including ChatGPT, for hours.
These outages highlight the challenges faced by OpenAI and its reliance on cloud infrastructure to power its services. As the adoption of AI tools continues to grow, ensuring robust uptime and minimizing disruptions will be crucial for maintaining user trust and satisfaction. OpenAI and Microsoft’s response to such incidents will likely be under scrutiny as they strive to provide seamless experiences for their users.
While Thursday’s outage has been resolved, the recurrence of similar issues underscores the need for improved resilience in the systems supporting ChatGPT and other AI applications.