DeepSeek’s AI chatbot went completely dark for over 10 hours this weekend, marking the longest service disruption since the Chinese startup’s R1 model took the world by storm in early 2025. The Information reports that both DeepSeek’s web and app-based chatbot services became unavailable starting Sunday evening, leaving millions of users without access.
What Happened
Users began reporting problems on Sunday night. DeepSeek’s own status page acknowledged an initial issue at 9:35 p.m. local time, then marked it resolved about two hours later. But that fix didn’t stick. A second wave of performance issues followed, and the service wasn’t fully restored until 10:33 a.m. Monday morning, stretching the total disruption well past the 10-hour mark.
DeepSeek offered no explanation for what caused the outage, which is consistent with the company’s typical silence on operational issues.
Why This Matters
This is significant because DeepSeek had maintained close to a 99% uptime record since its R1 model launched in January 2025. Until now, no previous outage had lasted longer than two hours. A 10+ hour blackout is a different category of problem entirely.
DeepSeek has rapidly become one of the most widely used AI chatbots globally, especially after its open-weight models shook the industry with their performance-to-cost ratio. An outage this long raises real questions:
- Infrastructure scaling: Can DeepSeek’s backend handle its explosive user growth? Caixin Global points to surging AI demand in China as a possible factor.
- Reliability for production use: Developers and businesses building on DeepSeek’s API need to know if this is a one-off or a sign of deeper infrastructure strain.
- Transparency: The company’s refusal to comment follows a pattern, but as more users and enterprises depend on the service, that silence becomes harder to justify.
What’s Behind It?
DeepSeek hasn’t said a word, but speculation among developers ranges from a system update ahead of a rumored new model release, to backend architectural changes, to simply being overwhelmed by a sudden user surge. This was DeepSeek’s second disruption in 60 days, which suggests the company may be hitting growing pains as demand outpaces its infrastructure buildout.
The Bigger Picture
Reliability is becoming the next competitive battleground in AI. OpenAI, Google, and Anthropic have all invested heavily in enterprise-grade uptime guarantees. DeepSeek disrupted the market on price and model efficiency, but enterprise customers and serious developers also need dependability. A 10-hour outage is the kind of incident that makes CTOs reconsider their AI provider choices.
What stands out here is the gap between DeepSeek’s model-level ambition and its operational maturity. Building a world-class model is one thing. Running it at scale with five-nines reliability is another challenge entirely.
For the full details, check out the original reporting from The Information and Bloomberg.