Microsoft confirmed on 16 June that it is routing GitHub traffic through Amazon Web Services — a remarkable admission that its own Azure cloud infrastructure cannot handle the volume of AI-generated code flowing through the world's largest code hosting platform.
The numbers are staggering. GitHub is processing 275 million commits per week, on pace for 14 billion commits in 2026 compared to 1 billion in 2025 — a fourteenfold increase driven almost entirely by AI coding agents. AI agent-opened pull requests grew from 4 million in September 2025 to 17 million by March 2026. The platform's infrastructure was designed for human-paced development workflows, not the continuous, parallel output of thousands of autonomous coding agents operating simultaneously.
The decision to route through AWS rather than scaling Azure capacity suggests the problem is architectural rather than simply a matter of adding more servers. GitHub's legacy systems were built around the assumption that individual developers would push code in discrete sessions. AI agents operate continuously, generating pull requests, responding to code reviews, and iterating on changes at machine speed — a fundamentally different usage pattern that strains every layer of the platform.
The irony is sharp: Microsoft owns both GitHub and Azure, yet neither can handle the traffic generated by the AI coding revolution that Microsoft itself helped accelerate through GitHub Copilot. The company is effectively paying its cloud competitor to keep its developer platform running.
For context engineers, GitHub's infrastructure crisis is a leading indicator of how AI agents will stress existing software infrastructure. Platforms designed for human-paced interaction — CI/CD systems, code review tools, package registries — will all face similar scaling challenges as AI agents become the primary producers of code.
