GitHub powering up four million extra cores to stop its service going down
Re-architects site to cope with AI slop-pocalypse
GitHub's Universe conference in San Francisco revealed significant infrastructure upgrades to combat soaring demand from AI agents. September 2026 saw 7.4 billion code commits, five times the volume recorded a year prior. This unprecedented traffic surge caught GitHub off guard, resulting in multiple outages and service disruptions.
AI-driven code commits began causing issues in December, leading to problems in February, April, and August, among other months. By May 7, the site's 90-day availability plummeted to around 84%, based on reconstructed data from GitHub's status page. In response, GitHub pledged to add four million CPU cores and 160 petabytes of high-speed storage to its infrastructure plans, along with enhanced compute capacity for GitHub Actions.
The company also aims to redesign its technical architecture to separate compute and storage, potentially increasing throughput by 35 times. Furthermore, GitHub plans to migrate to a multi-region deployment in Azure. Jay Parikh, EVP of CoreAI at Microsoft, highlighted these investments as essential for future growth. GitHub also introduced Copilot Fusion, a model routing service designed to optimize task completion rates and reduce token costs.
The service evaluates task requirements across various models and selects the most suitable approach. Additionally, GitHub unveiled Copilot Projects, Copilot Connectors, and Copilot Automations, and extended Copilot access to all GitHub Enterprise users at no additional cost. The company is also rolling out AI Scans to detect vulnerabilities in pull requests and new reachability analysis tools.
Written by urgent.news from The Register Science's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
