It was shortly before 10 p.m. China time when the first complaints began multiplying on social media. DeepSeek, the Chinese artificial intelligence chatbot that has gained 355 million users over the past fourteen months, stopped responding. No error message, no warning. Just silence and a blinking cursor.
A user with the nickname yezi888 summed it up quite aptly on the Chinese platform Xiaohongshu: “It was only when DeepSeek went down that I realized I no longer know how to work without it.” And she was far from alone.
More than seven hours of chaos
Let’s break down what actually happened. On Sunday, March 29, at around 9:35 p.m. local time, the company acknowledged the first issues on its status page. Two hours later, it announced a fix. But the relief did not last long. Shortly after midnight, the problems returned, this time even worse. DeepSeek engineers deployed fixes at 1:24 a.m. and then at 9:13 a.m. The issue was not fully resolved until 10:33 a.m. the following day. The total duration of the outage? Seven hours and thirteen minutes of continuous unavailability, as confirmed by outlets including Bloomberg.
For ordinary chatbot users, this was the longest outage since the R1 and V3 models catapulted DeepSeek into the ranks of the world’s most popular artificial intelligence services in early 2025. Until then, the web interface for end users had never experienced an outage lasting more than two hours.
What exactly caused such a massive collapse? No one knows. DeepSeek is sticking to its usual approach and has not released any explanation. It did not respond to media inquiries, including those from Bloomberg. The only thing users found on the status page was a message containing a typo: “Performance Agnormal” instead of “Abnormal.”
Experts have suggested several possible causes. It could have been a server failure, a bug following a software update, or infrastructure overload. DeepSeek has experienced similar problems before, but only with its API for developers, in late January 2025, when the service was facing a huge influx of new users.
Who benefited from the outage
While DeepSeek was down, its competitors celebrated. According to available information, traffic to alternative services rose significantly during the outage, particularly to Alibaba’s Qwen model. China’s artificial intelligence market is incredibly crowded right now. ByteDance, Tencent, and Alibaba have all introduced new models in recent months, and every opportunity to take a bite out of DeepSeek’s market share is welcome.
And this is where an interesting paradox lies. Since its debut, DeepSeek had built a reputation as a reliable service with nearly 99% uptime. People began to trust it so much that they integrated it into their workflows, business applications, and daily routines. But the more you rely on a single service, the harder the fall when it disappears.
Patch upon patch and no backup plan
Developer Ahmed M. Abdelfattah, who wrote about the outage on Medium, described his experience quite vividly. At three in the morning, he was going through the logs of his project connected to DeepSeek. Every query came back empty. No error, simply nothing. According to him, most of the 355 million users were unable to find an alternative quickly. And that is exactly the problem.
How many companies and developers have a backup plan ready in case their artificial intelligence provider goes down? Judging by the reactions on social media, the answer is: damn few.
The outage comes at an interesting time. Since the beginning of 2026, there have been rumors in expert circles that DeepSeek is preparing its next-generation model, probably called V4. The company itself has not said a word about it, but the timing is suspicious. Some commenters on Reddit speculated that the outage may have been related to a migration of server infrastructure ahead of the launch of the new model. None of this has been confirmed, however.
On the one hand, it was “only” a seven-hour chatbot outage. On the other hand, it is a warning to anyone who has entrusted their work, business, or creativity to a single service. For hundreds of millions of people, DeepSeek has become something like a digital backbone. And servers should not collapse without explanation in the middle of the night.
The company remains silent for now. The status page reports that all systems are operating normally. But the typo “Agnormal” in the official outage notice will probably remain for a while as an unpleasant reminder that even the most successful are not infallible.
Sources: reuters.com, timesofindia.indiatimes.com and tbsnews.net
