The two companies compete for cloud customers, for AI workloads and for the future of enterprise computing. Lately they are also trading favors. Microsoft has turned to Amazon Web Services for additional computing capacity to relieve pressure on GitHub, the code-hosting platform it bought in 2018, according to two people familiar with the matter.
The request follows a stretch of service interruptions on GitHub tied to its artificial-intelligence features, the people said. AI coding tools have flooded the platform with new code, and the surge has pushed GitHub’s own infrastructure past comfortable limits. Microsoft, which had planned to finish moving GitHub onto its Azure cloud by 2027, needed relief faster than its own build-out could provide.
The arrangement is unusual. AWS and Azure are direct rivals, and cooperation at this level is rare in an industry where switching suppliers is the usual remedy for capacity problems. For Amazon, the work has commercial logic: AWS sells compute to anyone, and a strained Microsoft is still a paying tenant. For Microsoft, the logic is starker: GitHub is a franchise too valuable to let slow down.
GitHub’s centrality is hard to overstate. More than 100 million developers use the platform, which hosts both the open-source commons and the private code of most of the software industry. Microsoft paid $7.5 billion for it in 2018, in a deal that shocked the developer community and later proved one of the company’s better bets: GitHub became the home of Copilot, the AI assistant that made Microsoft a leader of the coding-tools boom.
GitHub long ran on its own data centers, a legacy of its independent years, and its move to the cloud was always going to be delicate: the platform’s scale, and its role as the industry’s source of truth, made a botched migration unthinkable. The AI surge has made the migration urgent in a different way. Copilot now has tens of millions of users, and a growing share of the code pushed to GitHub is written by machines rather than people.
AI agents work around the clock, generating commits, pull requests and build runs at a pace no human team could match. The load spikes when agents run in batches, according to people familiar with the matter, and the platform’s systems were not built for that rhythm. For developers, the interruptions arrived at the worst time: agent-based tools run unattended, so a platform hiccup can corrupt whole batches of automated work. In the past, a developer at a keyboard could pause. An agent cannot.
The strain is a symptom of the industry’s wider condition. Every vendor selling AI productivity is also running the infrastructure that absorbs the output. Capacity planning, once a back-office discipline, has become a strategic constraint. Cloud providers are building data centers as fast as they can secure power, and the largest of them have begun rationing access to their newest chips.
GitHub is not alone. Rivals in the code-hosting market and the broader developer-tools industry are reporting similar pressure as AI agents multiply the volume of work pushed through their systems, according to industry executives. The difference is that GitHub sits at the center of the software world, so its outages are visible to everyone.
For Microsoft, renting from Amazon carries a cost beyond the bill. GitHub’s migration to Azure was meant to cut costs, deepen integration with Copilot and give Microsoft a flagship tenant for its cloud. Every month of AWS dependency is a month the migration timeline slips, and every outage on GitHub is a reason for its users to wonder whether the platform’s backbone is secure.
The arrangement also puts two of the industry’s most powerful companies in a strange position. Amazon and Microsoft have spent years poaching each other’s customers, and regulators on both sides of the Atlantic have scrutinized their dominance. A deal in which Microsoft rents compute from Amazon will be read as pragmatism or as dependence, depending on who is doing the reading.
Analysts said the episode shows how fast AI demand can outrun even the best-laid plans. Microsoft’s 2027 migration target predates the AI coding boom, and the boom has changed the math. The company could accelerate its Azure build-out, buy more capacity from other providers, or continue quietly renting from its rival while the migration catches up. None of the options is cheap; all of them are slower than the demand.
For GitHub users, the interruptions were a nuisance. For Microsoft, they were a signal about the platform’s future: the world’s largest code repository now depends on compute its owner does not fully control. The relationship with Amazon is transactional for now. Whether it stays that way depends on how quickly Microsoft can make Azure big enough to hold GitHub’s future.
