GitHub Grapples with Surging Traffic: Can Azure Keep Up?

GitHub's been in the spotlight for its ongoing battle with service availability. AI-assisted coding and agentic development workflows have spiked traffic, and GitHub's infrastructure is showing the strain. While the platform's tried expanding its capacity and offloading more workloads to Azure, reliability issues linger.

Expanding Capacity and Azure's Role

In its May 2026 Availability Report, GitHub reported nine incidents that degraded performance, slightly better than April's ten. That's a minor improvement, but not enough. Jakub Oleksy, SVP of software engineering at GitHub, admits there's a long way to go. 'We're making structural changes to permanently remove failure modes,' he stated.

GitHub had initially planned a 10x capacity increase back in October 2025. By February 2026, it was clear a 30x expansion was necessary. That's the kind of scale-up you expect when you're dealing with 1.4 billion commits a month, up from 1 billion for the entire previous year.

Infrastructure Strain: The Azure Dilemma

Microsoft's code hosting service is now serving 40% of monolith traffic from Azure, up from 8% in February. Git traffic's at 30%, and repository replication's nearly ubiquitous at 99%. But here's the catch: even Azure's been grappling with its own capacity problems. Is Azure the panacea GitHub hoped for?

Oleksy mentions isolating GitHub’s primary database cluster by segregating users, authentication, and authorization into separate domains. This should, in theory, prevent failures from cascading. Yet, GitHub's availability challenges persist.

Numbers Tell the Tale

The unofficial Missing GitHub Status Page tallies twelve incidents in May, with a 90-day uptime of just 87.26%. The monthly availability was 78.33% in April, 93.86% in May, and 88.39% so far in June. GitHub's own status page paints a far rosier picture, boasting figures around 99.9% uptime. These discrepancies raise a critical question: Which numbers do we trust?

If GitHub truly aims to be the backbone of agentic development, they'll need more than just Azure's promise. Slapping a model on a GPU rental isn't a convergence thesis. What's the real cost of inference when reliability is at stake? Without tangible improvements, GitHub might find itself at a crossroads, where trust in their infrastructure could wane as quickly as user patience.