GitHub Adds 4 Million CPU Cores as AI Agents Reshape Commit Volume

GitHub is scaling its underlying systems at a pace that shows how much AI agents are changing software development, not only in the code they write but in the volume of commits, pull requests and automated checks moving through the platform. The company has added four million CPU cores, 160 petabytes of high-speed storage and 50 terabits per second of network capacity during 2026 to handle rising activity, according to GitHub data reported by Xataka.

The infrastructure behind AI-generated code

AI coding assistants often get discussed in terms of what they can produce: a function, a bug fix, a finished feature. But every saved change, review request, automated check and synchronization step has to be recorded and coordinated. GitHub calls that less visible part of development increasingly important as agents work alongside human developers.

The new resources are additional capacity, not a measure of GitHub’s total infrastructure. They signal how much pressure the platform expects from a growing workload that includes both people and automated systems submitting changes more frequently.

Scale: commits jump from 12.9 billion to 34.3 billion

GitHub’s Octoverse 2026 data, as cited by Xataka, shows that annual commits across public and private repositories rose from 12.9 billion to 34.3 billion in the twelve months ending September 2026. In September alone, GitHub recorded 7.38 billion commits.

A commit is a saved version of changes to a project, not a line count. That distinction matters: the figures measure recorded activity, not how much code was generated or how large each change was. Still, the growth means GitHub must store, move and process far more frequent updates than before.

Agents change pull request work

More than 40 percent of pull requests in September 2026 began with a draft created by an AI agent that GitHub could detect. In March, that share was 17 percent. GitHub also found that new repositories with agent configuration files averaged eight times more commits than repositories without those files.

That association is notable but does not prove an equivalent productivity gain. Repositories without agent configuration files can still use AI, and the comparison alone does not establish code quality, review burden or delivery speed.

The strain is not only about storage

A single agent task can involve reading a repository, proposing changes, running checks and modifying the work again. The final result may be one improvement, but syncing that process with GitHub can require multiple commits, reads and automated validations. When many agents run in parallel, the platform has to handle a high frequency of small operations, not just larger codebases.

That workload pattern explains why capacity alone is not enough. GitHub has to process changes quickly enough to keep agents and human reviewers from waiting on the platform.

The August 17 outage exposed limits

On August 17, 2026, GitHub suffered an interruption lasting seven hours and 47 minutes that affected multiple services. The company said the incident began when traffic reached a new high and a critical component in a United States data center could not scale to meet demand. GitHub acknowledged that it had not expanded certain resources far enough in advance, but it did not directly attribute the outage to AI agents.

The event showed how quickly limits can appear when recorded activity spikes. It also helps explain why GitHub is now changing more than just raw capacity.

More than capacity: GitHub’s architecture shift

GitHub is migrating to a distributed infrastructure across multiple Microsoft Azure regions and working to separate storage from processing so each can scale independently. It is also trying to reduce operations that need to coordinate with each other.

In internal tests, the new architecture multiplied write capacity by up to 35 times, although that result does not yet represent general platform performance. The change is aimed at making repository infrastructure resilient to the faster automated workflow patterns becoming common with AI agents.

What this means for developers and platform teams

For development teams, the GitHub figures are a signal that agent-generated work can increase the number of commits, pull requests and CI/CD checks that infrastructure must absorb. Developers still need to review changes, verify they work and decide what to keep. The systems that turn those changes into working software are now part of the same scaling problem.

Several checks would add clarity: GitHub publishing more detailed Octoverse methodology, a follow-up incident report on the August outage, and evidence that the 35x write capacity result appears in a production service rather than internal tests.

Key facts

| Fact | Detail |
| GitHub infrastructure added in 2026 | 4 million CPU cores, 160 PB high-speed storage, 50 Tbps network capacity |
| Annual commits | 12.9 billion to 34.3 billion in twelve months through September 2026 |
| AI-drafted pull requests | More than 40% in September 2026, up from 17% in March 2026 |
| August 17, 2026 outage | 7 hours 47 minutes after traffic hit a new high and a critical US data center component failed to scale |

Source: Xataka IA, “GitHub ha tenido que hacer algo a gran escala para seguir el ritmo de la IA. Sus cifras explican hasta qué punto,” by Javier Marquez, published October 29, 2026. https://www.xataka.com/robotica-e-ia/github-ha-tenido-que-hacer-algo-a-gran-escala-para-seguir-ritmo-ia-sus-cifras-explican-que-punto

Datos clave