GitHub’s platform is absorbing 275 million commits per week, putting it on pace for 14 billion commits in 2026. That’s a 14x increase over all of 2025, when the platform celebrated hitting 1 billion total commits as a milestone. The source of the explosion: autonomous AI coding agents.
GitHub COO Kyle Daigle posted the numbers directly: “There were 1 billion commits in 2025. Now, it’s 275 million per week, on pace for 14 billion this year if growth remains linear (spoiler: it won’t).”
The Agent Surge in Numbers
The scale shift becomes clearest in pull request volume. According to The Information (as reported by Quasa), AI agents opened more than 17 million pull requests in March 2026, up from roughly 4 million in September 2025. That’s a 4x increase in six months.
GitHub Actions compute minutes tell a parallel story. Weekly compute minutes grew from 500 million in 2023 to 1 billion in 2025. In April 2026, a single week hit 2.1 billion minutes, as noted in Tomasz Tunguz’s analysis of the infrastructure demand curve.
Traffic Patterns That Break Assumptions
The problem isn’t just volume. AI agents interact with GitHub fundamentally differently than human developers. They clone repos, push branches, open PRs, run CI/CD pipelines, and iterate at machine speed across thousands of repositories simultaneously. They don’t browse the UI. They hit the API and CLI continuously.
Most of this agent traffic runs on free accounts and open-source repositories. GitHub’s generous public API limits and free-tier Actions minutes were designed for human developers and occasional CI bots, not for fleets of autonomous agents generating more code in an hour than an engineering team produces in a week.
Platform Strain Is Visible
The result is repeated outages and performance degradation tied directly to agent-driven traffic spikes, according to Quasa’s reporting. Daigle’s team is “pushing incredibly hard on more CPUs, scaling services, and strengthening GitHub’s core features,” but the growth curve has turned capacity planning into crisis response.
GitHub’s own product decisions confirm the pressure. InfoWorld reported last week that GitHub launched Stacked PRs specifically to help teams manage the AI-driven surge in code changes, breaking large agent-generated submissions into reviewable increments.
The Business Model Question
Every major AI coding tool, including Cursor, Claude Code, Devin, OpenDevin, and Windsurf, routes its output through GitHub. The platform has become the central nervous system of AI-driven software development by default.
That creates an unsustainable dynamic: the infrastructure serving 14 billion projected annual commits is being hammered primarily by agents that pay nothing. Agent-specific rate limits, dedicated AI agent plans with paid tiers, and usage-based billing for Actions are all likely on the near-term roadmap.
Daigle’s aside about linear growth is the key signal. The curve won’t flatten as more capable agents ship monthly. GitHub built itself into the rails of the AI coding ecosystem. Now it has to figure out pricing for passengers that never sleep.