Leading domestic AI company DeepSeek has seen its services return to stability after experiencing abnormal fluctuations for three consecutive days. Monitoring data shows that from March 29 to 31, DeepSeek's web-based chat, mobile app, and API interface all experienced varying degrees of connection interruptions or response delays.
According to the detailed records on the official status page, the durations of these three incidents were 1 hour and 48 minutes, 10 hours and 13 minutes, and 1 hour and 3 minutes respectively. The large-scale outage that occurred on March 30, lasting over 10 hours, had a significant impact on the developer ecosystem and enterprise applications. As of the time of this report, the official confirmed that all three incidents have been resolved, and the services have returned to normal operation.
Data statistics show that within the past 30 days, the overall availability metric of DeepSeek's web chat service was 98.61%. Although this figure is at the industry average level, frequent recent unexpected anomalies have triggered in-depth discussions within the industry about the capacity of large model infrastructure. Analysts believe that as the user base of large models grows explosively, the scheduling pressure of computing clusters and the resilience of network architecture have become new bottlenecks restricting the large-scale deployment of AI. This incident once again serves as a warning to the industry: while competing for model parameters and intelligent limits, building a foundational platform with high availability and disaster recovery capabilities has become an urgent priority for AI vendors.
