DeepSeek Goes Down for 7 Hours, Longest Outage Since Debut

DeepSeek went quiet this weekend. For about seven hours, users couldn't start new chats, and existing conversations just froze. The outage started around late Sunday night Beijing time and lasted into Monday morning. That's the longest stretch of downtime for the chatbot since it went viral in early 2025.

There were signs it briefly recovered before going down again, but the second fix took until morning. As usual with DeepSeek, no reason was given. Could have been servers acting up, could have been something in an update. They don't say.

Interestingly, the API, which developers use to build with DeepSeek, appears to have stayed up through all of this. The failure was in the Web Chat Service, the layer that handles logins, sessions, and getting users conversations in front of the actual model. So DeepSeek's brain was still working. It just couldn't get the words out.

DeepSeek has roughly 100–130 million monthly active users globally as of February, with most of that usage coming from China. That's a lot of people who woke up Monday morning to a chatbot that wouldn't talk back.