On October 15, 2026, the mainland Chinese artificial intelligence platform “DeepSeek” experienced service disruptions again. Both the website and mobile applications displayed errors, affecting some of the API services used by developers and businesses. The outage lasted nearly 5 hours, with many users reporting disruptions in their work.
According to DeepSeek’s service status records, the issue began at 8:11 am Beijing time. Users attempting to send messages were greeted with a “Service Unavailable” prompt on the webpage.
By 9:06 am, the company posted on the status page that they had identified the cause of the problem and were working on fixing it. By 12:49 pm, they reported that repairs had been made and they were monitoring the results, finally announcing service restoration at 1:04 pm.
Based on incomplete statistics from the service status records of DeepSeek by “First Financial,” the platform has experienced at least 10 significant outages this year, along with more frequent performance degradations. The most recent incident occurred on October 12, with partial interruption of web services for about 11 minutes; September 2 and 15 also saw interruptions in web and API services, with the latter lasting around 3 hours.
Although DeepSeek claims to have identified the cause of the malfunction, as of press time, they have not publicly disclosed specific details.
Tian Feng, the director of the artificial intelligence industry research institution “Think Tank Institute,” told “The Paper” that while DeepSeek previously limited the number of service requests, adjusting demand through peak and off-peak pricing to reflect its limited processing capacity. Sharing computational resources between new model testing and online services may further strain service resources.
Tian Feng also stated that it is currently uncertain whether this downtime is related to the release of a new version.
Another interviewed executive from an AI cloud computing service provider believed that this outage also reflects insufficient operational and maintenance infrastructure. Companies need to enhance testing, phased upgrades, traffic management, and fault rollbacks processes to alleviate the impact of resource constraints on services.
The outage occurred during the weekday morning peak working hours in mainland China. Regular users were unable to continue their conversations, and developers and businesses accessing DeepSeek through APIs were also affected. Some users commented, saying, “DeepSeek crashed… I’m in a panic.”
Tian Feng analyzed that the more tasks AI is involved in, the closer it is to a shutdown. After the interruption of tasks in progress, document modifications and tool invocation processes may need to be redone.
Later in the afternoon, DeepSeek issued an apology and compensation notice stating that they would provide account credits equal to the amount spent by API users from October 8 to 14, to offset future API charges. The company also pledged to investigate service stability issues and restructure to reduce single point dependencies.
Netizen “TechSister” expressed, “This morning when I opened DeepSeek, it crashed again…” She believed that large models have shifted from being “novelty tools” to essential work tools for many people, but service stability has not kept pace with the expansion. She also shared other netizens’ humor, stating, “The first thing to do when starting work is to see if DeepSeek can connect.”
Another netizen commented, “DeepSeek crashed, thankfully ChatGPT didn’t.” While another cautioned against excessive reliance on AI services, highlighting that there are multiple options available, stating, “If one crashes, there’s always another,” emphasizing the importance of not overly depending on AI.