When AI Demand Breaks the Forecast
Microsoft’s decision to use Amazon Web Services for GitHub shows how rapid AI-driven demand can outgrow even the best-planned cloud infrastructure, creating unexpected strain, outages, and emergency capacity deals between direct competitors. GitHub has become a core hub for AI-assisted coding, and that shift has changed its traffic profile faster than Microsoft’s Azure teams predicted. AI coding tools do not only help humans type code; they also trigger waves of automated commits, tests, and repository operations. Business Insider reports that GitHub commits were on pace to reach 14 billion in 2026, up from 1 billion in 2025, a level of growth that would stress any platform. This surge led to dozens of major outages in 2026, turning what looked like solid AI capacity planning into a visible GitHub infrastructure outage problem for everyday developers.
Calling a Cloud Rival: Inside the Microsoft–AWS Arrangement
According to Business Insider, Microsoft is adding extra computing capacity for GitHub through Amazon Web Services after AI-driven coding activity strained GitHub’s infrastructure beyond what Azure alone could handle. Microsoft still aims to move GitHub fully onto Azure by 2027, but the new reality is a multi‑cloud strategy: accelerate the Azure migration while using other providers when demand spikes. A Microsoft spokesperson said the spike in agentic development “has tested our infrastructure's limits” and confirmed that GitHub is tapping multiple cloud providers to gain “compute elasticity and horizontal scale.” Amazon, for its part, said customers choose AWS for reliable global infrastructure but declined to name Microsoft as a client. The move highlights how cloud infrastructure strain can override competitive positioning when service reliability and uptime are on the line.
Agentic Coding: From Autocomplete to AI Swarms
The core driver is not traditional autocomplete but agentic development, where AI systems act on code directly and repeatedly. A single coding agent session can read a repository, plan changes, edit files, run tests, and open pull requests, multiplying the load on GitHub’s compute, storage, and networking layers. Startup Fortune notes that Business Insider cited GitHub COO Kyle Daigle’s post that commits are tracking from 1 billion to 14 billion in one year, a step change that standard AI capacity planning did not anticipate. GitHub has leaned into this shift by tying Anthropic’s Claude and OpenAI‑based agents into GitHub, GitHub Mobile, and Visual Studio Code through Agent HQ for Copilot Pro Plus and Enterprise. Each new agent capability makes the platform more powerful—and more expensive and complex to keep online at scale.
Outages, Competition, and the Cost of AI Reliability
For developers, abstract talk of cloud infrastructure strain has become a daily pain. Business Insider reports GitHub suffered dozens of major outages in 2026; GitHub’s own availability update logged nine incidents of degraded performance in May alone. Mitchell Hashimoto of HashiCorp complained in April that GitHub was “no longer a place for serious work if it just blocks you out for hours per day, every day.” Those gaps have opened room for rivals like Cursor and Anthropic’s Claude Code, which market themselves on speed and reliability as much as features. Meanwhile, GitHub is adjusting its business model to match its new load profile: it is shifting all Copilot plans to usage‑based billing via GitHub AI Credits, calculated on token consumption, a clear signal that unlimited AI sessions and flat fees no longer line up economically.
What GitHub’s Struggles Reveal About Enterprise AI Scaling
Microsoft’s AWS move is a warning to every enterprise racing into AI: adoption curves can steepen faster than infrastructure and budgets. GitHub’s experience shows how AI capacity planning based on modest autocomplete tools can break when platforms shift to long‑running agents that behave more like swarms of junior developers. The result was a GitHub infrastructure outage pattern that forced an urgent multi‑cloud patch rather than a neat Azure‑only migration. Other tech giants are making similar moves, with SpaceX and Google agreeing to a deal in which Google will pay SpaceX USD 920 million (approx. RM4.28 billion) a month for AI compute capacity, even as Google Cloud sells compute to Anthropic. These cross‑cloud arrangements show that keeping AI services running now matters more than winning clean architecture diagrams or single‑vendor narratives.
