All posts
Cloud StrategyJune 18, 2026· 5 min read

Multi-Cloud Stops Being Optional: Microsoft Routes GitHub Traffic Through AWS

On June 16, 2026, Microsoft confirmed it is routing GitHub traffic through Amazon Web Services after a series of AI-agent-driven outages broke enterprise SLAs. When the largest cloud vendor in the world runs to its biggest rival, the multi-cloud debate is over.

On June 16, 2026, Microsoft confirmed a decision that would have been unthinkable five years ago: it is routing a portion of GitHub traffic through Amazon Web Services — its largest cloud competitor — following a series of AI-agent-driven outages that breached enterprise SLA commitments to GitHub's largest customers.

The reporting from TechTimes frames it as an AI-agent capacity crisis: Copilot-driven load, agentic build workflows, and a sharp rise in automated pull-request traffic overwhelmed capacity in ways that Azure alone could not absorb inside SLA windows. Whatever the exact root cause, the strategic signal is louder than the technical one.

What it means for the rest of us

If Microsoft is prepared to run GitHub — one of its crown-jewel acquisitions — partially on AWS to protect customer SLAs, the argument that 'we're standardizing on a single cloud for simplicity' has quietly lost. Multi-cloud is no longer a philosophical preference; it is a reliability primitive.

That does not mean every workload should be multi-cloud. It means the tier-1 workloads whose SLAs are contractual — payments, authentication, developer tooling, trading — should have a viable second path before they need one.

What a pragmatic 2026 multi-cloud looks like

  • One primary cloud per workload, with active-active only where the business case is real.
  • Portable data plane: object storage abstraction, database engines that run on both providers (Postgres, not proprietary), and Kubernetes as the compute substrate.
  • Identity, secrets, and observability standardized across providers so an operator can failover without relearning tools.
  • Contracts and cost models that assume you will move traffic — negotiated egress, not surprise egress.

The takeaway

The GitHub-on-AWS story is the clearest possible signal that hyperscaler concentration risk is now a board-level topic. Enterprises that architected for portability in 2024 and 2025 are the ones with options in 2026. Everyone else is one control-plane outage away from an all-hands weekend.