The number of commits on GitHub has doubled due to the use of AI – reaching 2.9 billion per month

The number of commits on github has doubled due to the use of ai reaching 29 billion per month

GitHub has published an analysis of the recent outage that left parts of the service down for over seven hours on Monday.

It wasn't a buggy deploy or a bad config — the root cause was plain capacity failure, i.e., systems couldn't keep up with the load.

Traffic reached a new peak, and a critical infrastructure component in our data center in the central United States was unable to scale along with it.

The surge was striking. GitHub reports monthly commits moving from 1.4 billion to 2.9 billion since April — more than double in roughly four months. That raw growth helps explain why some servers hit limits, but it doesn't make the outage any less jarring.

GitHub, on the reasons for the overload:

Monthly commits have grown from 1.4 billion to 2.9 billion since April. This growth explains the pressure on our systems but does not justify these outages.

Part of the story is AI-generated code. Big players have already said a growing fraction of code is produced by models: Google noted sizable AI-written portions of its codebase; Anthropic and OpenAI are pushing their own programming aids (Claude, Codex). In other words, more commits = more churn = more concurrent activity for services such as APIs, Actions, PR systems and Copilot. The status page has a deeper technical postmortem describing how simultaneous client activity cascaded into trouble.

Among the measures being taken, GitHub listed:

  • revising CPU and memory load alerts to proactively identify server components that could fail under peak traffic
  • unified rate limits and budgets for retry requests between services
  • variable timeouts to prevent "retry storms" and cascading load increases

Those fixes target the immediate failure modes, but they don't stop the underlying pressure. AI agents are committing code at machine speed, often nearly continuously, and GitHub is still the default hub for both commercial projects and hobby repos. So expect ongoing infrastructure strain as agent-driven activity spreads across the ecosystem — w/ unpredictable spikes that ops teams will have to chase.