Global leading AI large model service provider DeepSeek suffered a large-scale service outage starting in the evening of March 29th. This failure affected both the web and mobile app, with many users encountering "server busy" and network retry notifications during interactions, leading to interrupted conversations, data loss, and inability to log in, triggering a chain reaction. The incident quickly drew widespread attention and social discussion in the industry.

DeepSeek

Technical monitoring records show that the anomaly began at 9:35 PM on the 29th, and the official team immediately started investigating, announcing a preliminary recovery at 11:23 PM. However, early on the 30th, system performance experienced another fluctuation, and DeepSeek initiated a second investigation at 12:20 AM and implemented a repair plan at 1:24 AM. Despite multiple rounds of emergency interventions by the official team, as of 9:00 AM on the 30th, some functions had still not fully returned to normal. Currently, DeepSeek has not yet provided an official response regarding the underlying technical causes of this prolonged downtime or any subsequent compensation mechanisms.

This incident highlights the serious challenges that large model providers face in terms of computing power scheduling and backend architecture stability amid the explosive growth of user numbers.