AI agents are not just writing more code. They are stress-testing the infrastructure that stores, tests, and ships it.
GitHub reported 7.38 billion code commits in September 2026, more than five times the number from a year earlier. The platform attributes that surge directly to AI agents handling bug fixes and feature work in parallel. To keep pace, GitHub added 4 million CPU cores, 160 petabytes of high-speed storage, and 50 terabits per second of network capacity in 2026.
What the numbers actually mean
The commit count is only part of the story. Code pushes hit 3.35 billion per month in September, up 4.9 times year-on-year. Pull-request merges nearly quadrupled over the same period. GitHub Actions, the platform’s workflow automation service, ran 3.26 billion times in September, more than four times its year-earlier count.
Each push triggers a cascade of compute work: automated tests, security scans, CI checks, and version consistency verification. GitHub says a single push can generate thousands of additional reads by CI and code-scanning systems alone.

️ Why GitHub is rebuilding Git infrastructure
The platform’s existing architecture was designed around multiple data replicas spread across file servers. That worked fine at human-developer scale. At AI-agent scale, every extra replica involved in a write operation adds latency. GitHub’s October 2026 engineering update said the company is rebuilding its Git infrastructure to reduce unnecessary coordination between systems, separate storage from compute, and allow more operations to run simultaneously.
Heavy maintenance tasks, such as repository compaction and clean-up, are also being moved off the systems that handle live requests so background work stops competing with active code submissions.
The review problem no one is solving yet
More code means more pull requests means more review burden. GitHub introduced Copilot code review Max to help identify significant bugs before changes are accepted. The company is also migrating to a multi-region deployment on Microsoft Azure, with that work expected to finish in early 2027.
One thing GitHub was clear about: automated review tools do not replace human oversight. Developers are still responsible for understanding and approving the code that ships. Faster generation does not automatically mean better software.
