All Posts

Cloud & DevOps

Microsoft Rents AWS Capacity to Keep GitHub Running

Microsoft Rents AWS Capacity to Keep GitHub Running

Bhavika J

Techshorts Editorial Team

GitHub logged nine incidents of degraded service in May 2026 and dipped below 99 percent uptime for the month, according to the company's own availability report (GitHub, 2026). The worst single incident, on May 4, produced elevated latency and request failures across pull requests, issues, Actions, webhooks and Git operations for roughly an hour and six minutes, with knock-on effects on Codespaces, Pages, Packages and Copilot.

The strain traces to a specific cause: AI coding agents are opening far more pull requests than GitHub's infrastructure was sized for. Agent-opened pull requests rose from roughly 4 million in September 2025 to more than 17 million by March 2026. GitHub COO Kyle Daigle said in April 2026 that the platform was on pace for 14 billion commits in 2026, against 1 billion for all of 2025.

Business Insider reported on June 16, 2026 that Microsoft has started renting extra compute capacity from AWS to absorb the overflow, specifically for AI inference and training workloads tied to Copilot and AI-triggered GitHub Actions runs. Core version control, repository storage and authentication remain on Microsoft's own infrastructure. The report has been picked up and corroborated by outlets including TechRadar and Yahoo Finance.

The timing is awkward for Microsoft's own migration plan. GitHub has been moving its "monolith" traffic off older infrastructure and onto Azure. As of February 2026, though, only about 12.5 percent of that traffic had moved, against a target of 50 percent by July 2026 and full migration by 2027, per the same reporting. Microsoft has described the AWS arrangement as a stopgap rather than a change in that plan.

Who is affected

Enterprise customers who depend on GitHub Actions and Copilot uptime for CI/CD pipelines saw the direct impact of the May incidents. A separate, unrelated deadline affects a narrower group: administrators running GitHub Enterprise Server on premises. Starting August 18, 2026, GitHub will reject command-line support bundle uploads (via ghe-support-bundle, ghe-cluster-support-bundle or ghe-support-upload) from instances that are not on a current security patch (GitHub, 2026, changelog, July 22). The minimum required versions are 3.21.3, 3.20.5, 3.19.9, 3.18.12 or 3.17.18, or later.

What happens next

GitHub's June and July 2026 availability reports will show whether the AWS capacity addition actually stabilized uptime. The more consequential date is Microsoft's own July 2026 target of routing half of GitHub's monolith traffic through Azure. Whether that target was hit, and whether outages recur once it is, will be the clearer signal on whether this was a temporary capacity fix or a structural admission that Azure was undersized for the AI coding load GitHub now carries.

Sources

  1. GitHub, "GitHub availability report: May 2026" - https://github.blog/news-insights/company-news/github-availability-report-may-2026/
  2. GitHub, "Upcoming GHES change impacting uploading support bundles" - https://github.blog/changelog/2026-07-22-upcoming-ghes-change-impacting-uploading-support-bundles/
  3. TechRadar (reporting on Business Insider), "Microsoft forced to turn to AWS to boost GitHub cloud capacity following AI demand surge" - https://www.techradar.com/pro/microsoft-forced-to-turn-to-aws-to-boost-github-cloud-capacity-following-ai-demand-surge
  4. Yahoo Finance (Business Insider syndication), "Microsoft forced to turn to AWS to boost GitHub cloud capacity following AI demand surge" - https://finance.yahoo.com/technology/ai/articles/microsoft-forced-turn-aws-boost-142000306.html