On 15 October, DeepSeek's chatbot stopped answering for more than three hours, and the company's first public description of the problem was a "partial disruption". Users who were staring at an error message instead of a reply may have found that a generous reading.
What happened
The trouble began at about 08:11 local time. The report does not say which time zone that refers to, so treat the clock time with some care. The duration is clearer: the disruption ran for more than three hours.
Both ways into the service were hit, the web version and the mobile app. There was no workaround by switching devices.
What users saw
People trying to chat with the model were told, in effect, that the server was temporarily unavailable. The message appeared when they sent a prompt. Complaints centred on one point: they could not hold a normal conversation with the assistant.
What DeepSeek said
The company's status page first carried a notice of a partial disruption. Later, DeepSeek reported that it had identified the cause and that its engineers were working on a fix.
The company did not say what the cause was, at least in the material available here. "We found it" is a statement about progress, not about substance.
The capacity theory
Some industry analysts suggested the outage could stem from difficulty securing enough computing power. That is a plausible guess for a heavily used AI service, but it is a guess. The report states plainly that nothing official confirmed it, and DeepSeek has not tied the incident to compute shortages.
So the honest position is this:
- Confirmed: a long outage, affecting web and app, with a status page notice and a later statement that the cause had been found.
- Not confirmed: why it happened, and whether capacity played any part.
Where things stand
By the time the report was published, the app was back to working normally. There is no word on a detailed post-mortem, and none of the available material says one is coming.
For anyone who has built a workflow around a single hosted chatbot, the episode is a small, free reminder of what a single provider means. Three hours is not a catastrophe. It is, however, a long time to be told that a server is temporarily unavailable, and a longer time still to learn nothing about why.