GitHub is rebuilding its code storage as activity from AI agents grows
IA en un minuto newsroom · Editor: Jon Elgezabal
In 30 seconds
GitHub is overhauling the system that holds its repositories, pushed by a workload that AI agents have sent soaring in just one year. Data will no longer sit in complete copies spread across local disks but move to Azure Blob Storage, with lightweight machines handling reads. In its tests, the new design handles up to 35 times more writes.
GitHub is rebuilding the infrastructure that stores Git repositories, because activity from developers and AI agents on the platform has multiplied over the past year. The company explained it on its engineering blog on Tuesday, October 6.
In September, developers and agents made 7.38 billion commits on GitHub (each change saved to the code), more than five times as many as a year earlier. Pushes (sending those changes to the server) have reached 3.35 billion a month, up from 0.69 billion a year ago, 4.9 times more. Pull request merges (accepted change proposals) are at nearly four times their volume a year ago, and GitHub Actions, its automation service, ran 3.26 billion times in September, more than four times as many as a year earlier.
Overall, Git activity on the platform grew from 218.2 billion events a month to 473.3 billion between September 2025 and August 2026. The busiest repository saw roughly a billion requests in August. According to GitHub, an agent commits after nearly every action and waits for each push to complete before moving on, so latency a person would never notice becomes the limit.
Until now, each repository has been stored by a system called Spokes, with a full copy on the local disks of several servers, five by default. The new architecture separates storage from compute. Repository data lives in Azure Blob Storage, Microsoft's cloud storage service, and lightweight servers act as a cache for reads. In internal benchmarks, it has delivered up to 35 times higher write throughput.
GitHub says it is making the change while the service keeps running, without asking anyone to change how they build software. It has not given a date for completing it.
Why it matters · analysis and opinion
Until recently, a service like GitHub was sized with people in mind, who save changes every so often. An agent works differently: it saves after almost every step and waits for the server to confirm before carrying on, so any delay piles up. That GitHub has to rebuild its foundation for this reason shows how much work development teams are already handing to these programs. For people using the platform, the promise is that day-to-day work does not change. Where it may show most is in projects with heavy continuous integration and several agents at once, exactly the workloads the company itself singles out as the most demanding. What is still missing is when the change will reach every repository, because there is no timeline.
Official source: GitHub · Written with the help of AI: how we make the news
