DeepSeek hit by five-hour outage, offers partial refunds

DeepSeek suffered a roughly five-hour outage that knocked out its chat service along with two of its models, the latest in a string of disruptions to hit the Chinese AI developer this year.

The outage ran from the morning into early afternoon, during which users were unable to reliably access the affected services. DeepSeek issued partial refunds to subscribers whose access was disrupted but did not publicly identify what caused the failure.

The company has not confirmed a root cause, but the outage has fed speculation that it may be linked to constraints on computing capacity, as US export restrictions on advanced chips continue to limit the processing power available to Chinese AI firms. DeepSeek has previously drawn attention for building competitive models despite restricted access to the most advanced hardware, a feat that has also left it more exposed to capacity crunches when demand spikes or infrastructure falters.

Repeated outages carry real reputational risk for a company that built its profile on offering a cheaper, more efficient alternative to Western AI labs. Each disruption gives enterprise customers evaluating reliability alongside cost another data point to weigh, even as DeepSeek continues to attract a large user base drawn by its pricing and performance.

The company has not said whether it plans any structural changes to its infrastructure to reduce the likelihood of future outages, and has offered no timeline for a full explanation of what went wrong this time.