GitHub Reports Increase in AI-Generated Code Usage

GitHub, the world's largest code repository, added 4 million CPU cores, 160 PB of storage, and 50 Tbps of network bandwidth over the year to handle the influx of AI-generated code. This is stated in the company's technical report. According to the report, the surge is driven by agent-based AI, and the platform had to not only increase capacity but also rethink its technological infrastructure.

Activity in repositories has sharply increased: commits, which are operations of adding and deleting code, have increased sixfold, and merged pull requests—requests for merging that record a new feature or fix—have quadrupled. The growth rate is accelerating: in April, the platform had 1.4 billion commits per month, in August—2.9 billion, and in September—already 7.4 billion. Over the past 12 months, GitHub recorded 17.9 billion pushes—more than in the entire years of 2024 and 2025 combined.

According to GitHub's estimate, the surge is driven by agent-based AI. By September, more than 40% of pull requests were initiated by agents compared to 17% in March 2026. Repositories launched with AI generate approximately eight times more activity than those created manually.

Such a flow of code creates a disproportionate load on CI/CD pipelines. With each commit or pull request merge, numerous checks are triggered: searching for unfinished tasks, linting for code uniformity, code tests, and integration tests. The latter are necessary to ensure that new code does not break existing application functionality and works correctly with related programs, services, and APIs. According to GitHub, this load required both modernization and a reevaluation of the platform's infrastructure.