GitHub Runs on AWS Now: How AI Coding Broke the Platform

GitHub Is Running on AWS. Here's Why AI Coding Broke It.

GitHub commits are on pace to hit 14 billion in 2026. In 2025, the total was 1 billion. That is not organic growth. It is AI agents committing code around the clock — and it has pushed a platform built for human developers to its limits.

The fallout: 19 service-degrading incidents across April and May. Agent session wait times spiking to 54 minutes. And Microsoft, which owns both GitHub and Azure, quietly contracting Amazon Web Services to keep the platform running, according to byteiota.com's infrastructure breakdown.

The Numbers Behind the Crisis

Claude Code alone accounts for 2.6 million commits per week. In September 2025, that number was roughly 100,000. That is a 25x increase in six months. It now represents 4.5% of all public commits on GitHub.

AI agent pull requests tell the same story. They grew from 4 million per month in September 2025 to 17 million by March 2026 — a 325% increase in six months.

GitHub's infrastructure team planned for a 10x capacity expansion in October 2025. By February 2026, they revised that to 30x. GitHub SVP Jakub Oleksy stated: "We're now serving 40 percent of monolith traffic from Azure, up from 8 percent in February."

Why GitHub Needed AWS When It Already Has Azure

Microsoft owns Azure. It has been migrating GitHub to Azure since the 2018 acquisition. That migration is real and ongoing. But AI traffic outpaced the migration timeline. Azure could not absorb the load fast enough, so GitHub contracted AWS for overflow. GitHub calls this "multi-cloud resilience."

You might also like

DOJ Backs xAI in Clean Air Act Suit, Cites National Security

AWS WAF Now Lets Publishers Charge AI Bots

France Drops Palantir, Pledges €655M for Sovereign AI

Goldman Sachs: AI "Buy Everything" Era Is Over

GitHub CTO Vladimir Fedorov described the Azure migration as "existential" in an internal memo. He wrote: "We have to do this. It's existential for GitHub to have the ability to scale to meet the demands of AI and Copilot, and Azure is our path forward." He also acknowledged that earlier Azure migration attempts had "dragged on" and failed, per techbuzz.ai's reporting on the migration.

GitHub COO Kyle Daigle confirmed the migration to The Verge: "We need to scale faster to meet the explosive growth in developer activity and AI-powered workflows, and our current infrastructure is hitting its limits."

What the Worst Outages Looked Like

April was the worst month. GitHub logged ten incidents. Session start failure rates hit 84%, peaking at 97.5%. Agent wait times reached 54 minutes. One cascade took code search down for 8.7 hours, degraded Copilot's cloud agent for four hours, and knocked the Copilot backend offline for 2.7 hours — all at the same time.

May had nine incidents. A data integrity failure on April 23 affected 2,092 pull requests. A bug caused squash merges to drop commits when a merge queue held more than one PR. Companies including Modal and Zipline were hit. Customers had to manually restore missing code from git history, as The Pragmatic Engineer reported.

Can Duruk, a software engineer at Modal, said: "The COO going out of their way to find a huge denominator to make the impact appear small feels very dishonest; versus a sincere apology about how this invalidates their entire promise to their customers."

The official GitHub status page reports roughly 99.9% uptime. A third-party tracker called the "Missing GitHub Status Page" put the 90-day figure at 87.26%. Monthly availability ranged between 78% and 94%.

Why AI Traffic Breaks Platforms Differently

A human developer commits a few times a day. They open a pull request, wait for CI, then respond. That is a slow, sequential workflow.

An AI agent commits continuously. It triggers downstream processes the moment each change lands. One AI agent PR does not create one unit of compute load. It spawns a CI pipeline. That triggers webhooks. Those activate code review bots. Those generate Copilot suggestions. Those create more commits.

Open source maintainer Xavier Portilla Edo estimated that only 1 in 10 AI-generated PRs is legitimate. The human review burden exploded even as human commit rates stayed flat.

We see this pattern playing out across the industry. As AI agent usage scales, infrastructure built for human-paced workflows faces the same mismatch. The Claude Code surge is one concrete data point in a broader stress test. Even the economics are shifting — rising AI token costs are part of the same pressure.

What GitHub Is Doing Now

GitHub has set a clear priority order: availability first, then capacity, then new features. Concrete steps include:

• Isolating Git and GitHub Actions from other workloads
• Accelerating the Azure migration, with a 12-month core migration target and a full data center exit planned within two years
• Using AWS for overflow capacity
• Launching usage-based billing on June 1, 2026
• Pausing new sign-ups for Pro, Pro+, and Student Copilot plans while rebalancing capacity

The full exit from GitHub's own Virginia-based hardware is planned within two years. For teams building on AI-native workflows or agentic coding tools, GitHub's capacity constraints are the confirmed near-term factor to plan around, per bingx.com's reporting on the AWS contracts.