GitHub Rebuilds Git Repository Architecture for AI Agent Traffic
GitHub says internal tests of its new repository architecture showed 35 times faster write performance, as it redesigns storage from scratch.
GitHub is redesigning its Git repository architecture from scratch in response to a surge in traffic driven by AI agents, according to digitaltoday.co.kr. Internal tests of the new architecture showed write performance 35 times faster than the current system.
The company said developers' workflows, code reviews and security settings will remain unchanged. Only the internal architecture behind them is being altered.
Under the new design, data is written only once to Azure Blob Storage, which then handles replication and backups instead of GitHub maintaining multiple full copies.
GitHub did not disclose a timeline for the transition.
Pressure on the existing system
The current architecture, known internally as Spokes, stores multiple full copies of a repository across disks and requires several copies to confirm a write. That approach supports reliability but limits speed.
From September 2025 to August 2026, monthly events on GitHub more than doubled, from 218.2 billion to 473.3 billion. Commits reached 7.38 billion in September, five times more than a year earlier.
The service also logged 10 performance-degrading incidents in April and 9 in May.
Quick answers
How much faster is GitHub's new repository architecture?
GitHub said internal tests showed 35 times faster write performance.
Will developers notice changes to their GitHub workflows?
GitHub plans to keep workflows, code reviews and security settings unchanged, altering only the internal architecture.
When will GitHub complete the transition?
GitHub did not disclose a timeline for the transition.