(Yicai) Oct. 16 -- DeepSeek, one of China’s artificial intelligence champions, has apologised for an application programming interface outage that lasted nearly five hours yesterday and announced compensation for affected users.

The compensation credited to each user account will depend on its API usage and spending during the week of Oct. 8 to 14, DeepSeek said yesterday. The amount varies from a few Chinese cents to more than CNY1,000 (USD149), according to social media posts shared by users.

The outage lasted from 8.11 a.m. to 1.04 p.m. It was one of DeepSeek’s more serious service disruptions this year, occurring during peak working hours and likely causing substantial inconvenience for developers and businesses that rely on its API.

The Hangzhou-based company said it had opened an investigation into service stability and was implementing thorough corrective measures. These include identifying and eliminating single points of dependency to prevent similar incidents from recurring.

Some AI industry analysts believe that yesterday’s outage may have stemmed from a shortage of computing capacity supply.

DeepSeek's official service status records show indicate that it has experienced at least 10 major service disruptions so far this year, ranging from service outages to degraded performance. A prolonged disruption happened on Sept. 15, when DeepSeek's web platform and API services were partially unavailable for about three hours.

Privately held DeepSeek stormed onto the global AI stage in January last year with the release of its ChatGPT-style R1 large language model, which paired strong reasoning performance with dramatically lower running costs. Internet pioneer and Silicon Valley venture capitalist Marc Andreessen famously described it as an “AI Sputnik moment.”

Editor: Futura Costaglione