China’s answer to OpenAI’s ChatGPT recently suffered its longest-ever outage since its public debut. DeepSeek reportedly went out of over 7 hours, disrupting services across China and causing the Chinese social media to explode with memes and jokes. While the company did not reveal the reasons behind the chatbot’s outage, it managed to keep its entire userbase updated with the live status and eventually fixed it the next day.
The outage began around 9:35 pm local time on Sunday, March 29, and lasted until approximately 10:33 am on Monday, March 30 — spanning a total of 7 hours and 13 minutes. Users across China reported an inability to access the web and app versions of the chatbot, sparking widespread complaints on social media. DeepSeek’s official status page marked the incident as a “major outage” and deployed multiple system updates to restore services.
Why was DeepSeek out?
DeepSeek’s status page showed an initial issue acknowledged late Sunday evening, followed by a temporary resolution. However, problems recurred in the early hours of Monday, requiring further interventions. Full services were eventually restored by late morning, with the company stating it was continuing to monitor the platform for stability.
This marks the most significant downtime for DeepSeek since its R1 and V3 models gained massive popularity early last year, the ones that went on to briefly challenge global leaders like ChatGPT. The company, however, has not yet publicly disclosed the exact cause of the outage, even after it involved multiple rounds of investigation and fixes during the night.
Rumours, however, hint at DeepSeek readying its servers for updating the backend for a new DeepSeek V4 model, which is expected to be released in April 2026. Some even speculate the possibility of a server overload owing to the older AI chips the Chinese firm has to rely on.
Is DeepSeek back online?
As of now, DeepSeek’s chatbot and API services are reported to be fully operational again. The company has not issued a detailed post-mortem statement on the root cause.
Making a public debut in early 2025, DeepSeek has been considered as one of the most formidable challengers in the global AI space. Models like the DeepSeek-V3 and R1 have been found to be cost-efficient yet almost as capable as the leading AI models from the global players like Google, OpenAI, and Anthropic.
What’s even more interesting is that, unlike the competition, DeepSeek models are open-source and can be used by anyone. The current-gen DeepSeek V3.2 model competes with GPT 4.5 and GPT 5 in terms of coding and reasoning, but falls behind Anthropic’s latest Claude models.
DeepSeek, however, intends to release the DeepSeek V4 model in April 2026, with critics expecting the model to compete with the latest releases from rivals like Google’s Gemini 3, OpenAI GPT-5.4 and Anthropic Claude Opus/Sonnet 4.6.