GitHub outage traced to capacity limits as commits nearly double

GitHub's August 17 outage lasted nearly eight hours, from 13:28 to 21:15 UTC, and knocked out Issues, pull requests, the API, Actions, and Copilot, according to a postmortem GitHub published this week. The company says the cause wasn't a code or configuration change, it was a straightforward capacity failure: a data center component overwhelmed by traffic it wasn't sized for.

The scale context is the real story. GitHub's own numbers show monthly commits nearly doubling since April, from 1.4 billion to 2.9 billion, alongside 130 million merged pull requests and 24 million new repositories a month, growth GitHub attributes largely to AI coding agents committing code at a pace human-only workflows never produced. Trailing 90-day uptime on GitHub Actions fell to 99.33% as a result, burning through most of a year's downtime budget in one afternoon.

GitHub's fix so far is mostly brute-force capacity: more than 3 million additional CPU cores, 120 petabytes of high-speed storage, and an accelerated migration to Azure, which now handles roughly 58% of platform load and half of all Git operations, up from 12% in May.

For a small team running CI/CD on GitHub Actions, this is the second capacity-driven outage in under two weeks, not a one-off. Nothing about a minimal pipeline gets corrupted when GitHub goes down, deploys just stall, but the growing frequency argues for treating a GitHub Actions outage as a real, budgeted failure mode rather than a rare edge case.