Ex-Microsoft engineer believes Azure problems stem from talent exodus

When federal cybersecurity experts in 2024 described the Microsoft 365 Government Community Cloud High (GCC High) service as "trash," the tech industry held its breath. Although the phrasing they used was much more blunt, the core of the problem remains unchanged: the foundations upon which Microsoft's cloud is built are beginning to crack. Axel Rietschin, an engineer with many years of tenure in Redmond, sheds new light on this situation, suggesting that Azure's current problems are not just a matter of technology, but primarily the result of years of neglecting human capital and a mass exodus of talent.

Rietschin, who spent eight years as a Windows Base Kernel engineer and a year in the Azure Core Compute team, published a series of essays outlining a picture of a platform built in haste and under immense competitive pressure. According to him, Microsoft, in an attempt to catch up with Amazon Web Services (AWS) in 2008, prioritized speed of deployment at the expense of stability and long-term architectural vision. What the world sees as a powerful cloud infrastructure resembles, from the inside, a system constantly kept on life support by exhausted engineers fighting growing technical debt.

Foundations on Sand and the Chase After Amazon

The history of Azure is a story of "wishful thinking" regarding the pace of the platform's stabilization. Rietschin argues that the rushed start forced design decisions that are still causing issues today. A lack of discipline in software testing and insufficient quality control became the norm rather than the exception. Instead of building a solid structure, Microsoft focused on marketing and aggressive scaling, leading to a situation where the service never functioned as independently and smoothly as customers were promised.

However, the most dangerous phenomenon pointed out by the former engineer is "knowledge dilution" caused by high employee turnover. When experienced architects leave the company, they take with them a unique understanding of critical systems. New staff, often lacking proper mentoring, are unable to effectively manage such a complex organism. This leads to a vicious cycle: the system generates errors, engineers are overloaded with fixing them, which in turn prompts them to quit, further weakening the team.

Rising frustration within Microsoft coincides with increasingly loud signals of dissatisfaction coming from the outside. Reports published by ProPublica detail the US government's problems with Azure services. However, the most telling warning sign was the decision by OpenAI on March 10, 2025. The AI giant entered into a deal for computing power worth $11.9 billion with the company CoreWeave. Rietschin interprets this unequivocally: it is a vote of no confidence in Azure. If Microsoft's most important partner is looking for infrastructure elsewhere, it means Azure is unable to meet the demands of scale and performance even for its key projects.

AI as a Catalyst for the Staffing Crisis

The current fascination with artificial intelligence, instead of helping to solve infrastructure problems, seems to be deepening them. Microsoft, like many other corporations, succumbed to the illusion that AI would allow for workforce reductions while increasing productivity. The result of this thinking was a wave of layoffs – between May and July 2025, approximately 15,000 people left the company. Such a strategy, according to Rietschin, is short-sighted. AI does not replace engineers; it generates new challenges that require even more human attention.

• Mass production of code: AI agents, such as Claude Code, are capable of generating tens of thousands of lines of code in a short time, which drastically increases the load on CI/CD (Continuous Integration/Continuous Deployment) processes.
• Increased demand for power: Every commit generated by AI must be tested, deployed, and handled by database servers, which intensifies the "computational hunger."
• Low code quality: Code generated by AI is often based on statistical probability rather than an understanding of system logic, leading to subtle, hard-to-detect bugs.

We are currently observing a phenomenon that Martin Alderson, co-founder of catchmetrics.io, calls the "coming compute crunch." AI not only consumes massive resources for training models but generates a giant secondary load. Data from the Claude’s Code service shows a fourfold increase in the number of commits by Anthropic AI agents in just three months. This flood of code is overwhelming infrastructure that is already barely coping with current traffic.

GitHub Instability and Risky Migrations

Azure's problems are ricocheting into other services in the Microsoft ecosystem. Unofficial reports indicate that the availability of the GitHub platform fell below 90%. Interestingly, as a solution to these problems, GitHub CTO Vlad Fedorov pointed to... migration to Azure. According to his announcements, by July 2025, 50% of GitHub traffic is to be handled by the Azure Central US region. However, skeptical voices are appearing in the industry – is moving an unstable service to a platform struggling with its own structural problems really the path to "global resilience"?

Rietschin maintains distance from these speculations, noting that it is difficult to unequivocally link GitHub outages to Azure without insight into internal data. Nonetheless, the timing of these events is concerning. For many observers, the attempt to "simplify architecture" by moving to Azure managed services looks like an attempt to escape forward at a time when foundations require a thorough overhaul, not another layer of abstraction.

"LLM models are great at reproducing patterns, but their ability to find bugs stems from observing deviations from statistical expectations, not from an actual understanding of the code. I do not share the optimism regarding the replacement of engineers by AI," emphasizes Axel Rietschin.

The vision in which artificial intelligence independently writes, tests, and repairs software seems distant for now. In reality, the tech industry faces a paradox: the more code AI generates, the more experienced people are needed to verify it and maintain the infrastructure on which that code runs. Microsoft, by choosing mass staff cuts during the AI boom, may permanently damage its ability to maintain Azure as a trusted enterprise-grade platform.

Investing in People as the Only Way Out

Rietschin's analysis leads to one key conclusion: Microsoft must stop treating engineers as a cost and start seeing them as the most important resource guaranteeing stability. His recipe for healing Azure is simple in theory but difficult for a margin-optimized corporation to accept: a return to mentoring, training staff at all levels, and winning back the experienced technical leaders who have left in recent years.

The current Azure situation is a warning for the entire technology sector. Excessive trust in Copilot systems or other programming aid tools, while simultaneously getting rid of the "human failsafe," leads to a degradation of service quality. If Microsoft does not change course and reinvest in people, Azure could become a symbol of an era where technology outgrew its creators, and the lack of hands to work on "keeping the platform alive" led to its ultimate failure. In a world where the cloud is the backbone of the global economy, such a prospect is more than unsettling.