AI Coding Agents Push GitHub Past Its Limits

GitHub is straining under a workload that looks nothing like the traffic it was built to handle. Through 2026, autonomous coding tools have hammered the platform around the clock, and the numbers tell the story. According to GitHub CTO Vlad Fedorov, overall load on the platform has climbed roughly 3.5 times over the past two years, with much of that growth coming from AI agents that push code continuously rather than only during normal working hours.

The scale of automated activity is hard to overstate. Claude Code by itself is responsible for about 2.6 million commits every week. Pull requests opened by AI agents have exploded as well, jumping from roughly 4 million per month in September 2025 to about 17 million by March 2026 — a more than fourfold increase in just half a year. That kind of relentless, machine-driven demand is what has tipped GitHub's infrastructure into crisis.

Outages Mount as Availability Slips to 85%

The pressure has translated into real downtime. Across April and May, GitHub logged 19 separate incidents that degraded service, and independent monitoring put the platform's availability as low as 85 percent over a 90-day window — a level that makes dependable, full-day development difficult to sustain.

The reliability problems have started to cost GitHub credibility with high-profile developers. Mitchell Hashimoto, the founder of HashiCorp and the creator of Ghostty, publicly walked away from the platform out of frustration, arguing that it had stopped being a viable home for serious work when it locks users out for hours at a time, day after day.

Why Microsoft Is Buying Capacity From a Direct Rival

To keep the platform usable, Microsoft has begun purchasing cloud capacity from Amazon Web Services — a direct competitor — according to people familiar with the plans. The decision stands out precisely because Microsoft has been working to move GitHub entirely onto its own Azure cloud, which makes any turn toward AWS an unusual concession.

Microsoft has positioned the move as part of a wider multi-cloud approach rather than a retreat from Azure. In a statement, the company said it is simultaneously accelerating its migration to Azure and continuing to explore a multi-cloud strategy to keep pace with growth. In practice, that means leaning on AWS to absorb overflow demand while the larger Azure transition continues.

The Azure Migration That Demand Outran

GitHub started shifting off its own data centers and onto Azure in October 2025, a move originally expected to take about a year. The problem was that AI-driven demand grew faster than the migration could proceed. When the project began, GitHub's infrastructure team was planning for a tenfold capacity increase; by February 2026, they had concluded they actually needed to build for 30 times current scale.

Even so, the shift toward Azure is well underway. GitHub SVP Jakub Oleksy reported that about 40 percent of the platform's monolith traffic now runs on Azure, up sharply from just 8 percent in February. The capacity purchased from AWS is meant to bridge the gap while that transition finishes.

A Pragmatic Concession on Cloud Loyalty

The arrangement is a clear illustration of how quickly corporate cloud allegiance gives way when uptime is on the line. GitHub has spelled out its priorities in order — availability first, then capacity, then new features — and using a rival's infrastructure for overflow is the practical expression of that ranking.

The company has made other adjustments to steady the platform as well. On June 1 it moved to usage-based billing, and it paused new individual plan sign-ups so it could rebalance capacity toward existing users while it works through the crunch.