Story perspectives

DeepSeek Refunds API Users After Nearly 5-Hour Outage

By Drooid · Oct 15, 2026, 11:08 PM UTC · How we work

Summary

DeepSeek Refunds Users

DeepSeek AI refunded all users for API usage after a five-hour outage.
The outage ran from 8:11 a.m. to 1:04 p.m. Beijing time on Oct 15 2026, lasting 4 h 52 m.
Refunds covered spending from Oct 8-14, ranging from a few yuan to over 3,000 yuan ($449).
DeepSeek apologized and will investigate to eliminate single-point dependencies after 13 incidents since September, per Caixin.

Summary

Reddit Links Outage

Reddit users linked the outage to a rumored DeepSeek-V5 model release.
A new Bilibili account called DeepSeekHarness attracted over 10,000 followers during the disruption.
StatusGator recorded 86 fault reports within 24 hours, showing widespread failures.
User-reported issues were 57% app problems, 41% website issues, and 2% login failures.
Users in Luxembourg, Poland and New York posted “server temporarily unavailable” messages, highlighting global impact.

History

Historical Context

In November 2017, Amazon Web Services experienced a major outage that halted compute and storage in its US East-1 region, affecting thousands of customers. The event, occurring just 15 years after AWS launched in 2002, underscored single-point vulnerabilities that AI firms such as DeepSeek now reference when urging diversified cloud infrastructure.

1 of 3

DeepSeek Refunds Users

DeepSeek AI refunded all users for API usage after a five-hour outage.
The outage ran from 8:11 a.m. to 1:04 p.m. Beijing time on Oct 15 2026, lasting 4 h 52 m.
Refunds covered spending from Oct 8-14, ranging from a few yuan to over 3,000 yuan ($449).
DeepSeek apologized and will investigate to eliminate single-point dependencies after 13 incidents since September, per Caixin.

1 / 3