For a significant portion of the past year, GitHub has deviated from the reliable utility that developers have come to depend on. A concerning pattern of outages, affecting critical services such as search functionality, GitHub Actions, and the crucial CI/CD pipelines that rely upon them, has become an uncomfortably regular occurrence. The platform has logged hundreds of incidents over the preceding twelve months, prompting public apologies from the company. In an exclusive interview with The New Stack, GitHub Chief Operating Officer Kyle Daigle, who also holds the position of CMO of Developer at Microsoft, addressed the underlying causes of these disruptions and outlined the comprehensive strategies GitHub is implementing to rectify the situation.
The Unforeseen Surge: Why 100% Growth Projections Fell Short
GitHub has been transparent about the extraordinary growth it is currently experiencing, a phenomenon driven by the burgeoning era of agentic coding. This surge in demand has far surpassed even the most ambitious projections for hyper-growth cloud companies. Daigle emphasized that the current situation necessitates an "all hands on deck" approach for the entire organization.
"It’s not merely about implementing standard scaling solutions typical in the cloud era, such as acquiring larger or more numerous machines," Daigle explained to The New Stack. "We are now focused on ensuring we can scale at rates of 30 to 40 times over the course of the next year, a significant leap from the historically impressive 100 percent year-over-year growth we previously planned for."
Daigle admitted that GitHub had anticipated a surprise if growth escalated from 50 percent to 200 percent. "It turns out that was a conservative estimate, and the actual requirement is far greater," he stated. The engineering team is now working diligently to enhance the system’s capacity to handle an influx of commits, pull requests, and issues that is 30 times the current volume.
This situation carries a degree of irony, given that GitHub itself played a pivotal role in catalyzing this AI-driven development wave with the 2021 launch of Copilot. Copilot effectively evangelized AI code generation among developers, fostering a reliance on these tools. Now, GitHub is experiencing the strain of this widespread adoption. In 2025 alone, the service processed an astonishing 1 billion commits, a figure that has now escalated to 1.4 billion per month. Daigle revealed that AI agents alone are generating over 17 million pull requests monthly.
Deconstructing "All Hands on Deck": A Multi-faceted Approach to Stability
To accommodate this unprecedented demand, GitHub has been actively migrating its infrastructure from its own data centers to Microsoft’s Azure cloud. However, Daigle stressed that this transition is about more than just increasing raw capacity.
"Our primary focus has been on scaling not only through conventional means, such as acquiring more CPUs and implementing standard horizontal and vertical scaling techniques, but more critically, on deeply investigating and updating our underlying systems," Daigle elaborated. "This involves rebuilding or enhancing the often-hidden systems that perform the core functions of our platform."
Early efforts concentrated on alleviating pressure on the database, as highlighted by GitHub Chief Technology Officer Vlad Fedorov in a blog post earlier this year. GitHub successfully addressed MySQL contention issues, migrated webhook processing entirely off MySQL, and redesigned its session cache and authentication workflows to reduce database load.
For GitHub Actions specifically, Daigle indicated that the dispatching mechanism for jobs to runners underwent a complete rewrite. The overarching architectural objective is to create a robust isolation strategy, segregating critical services like Actions and Git from other components. This ensures that a struggling subsystem does not cascade failures to other parts of the platform. Furthermore, GitHub is actively moving performance-sensitive code from its Ruby monolith into Go, a language known for its efficiency and concurrency.
"We have effectively addressed most of the readily achievable improvements," Daigle stated, acknowledging that the tangible results of these enhancements are often difficult to quantify. "It’s the classic catch-22 of improving availability; when everything is functioning smoothly, it’s challenging to explicitly demonstrate the impact of specific improvements."
GitHub is also leveraging the extensive resources and expertise of Microsoft. "It is truly an ‘all hands on deck’ situation here at GitHub," Daigle reiterated. "We are receiving unprecedented support, with experienced engineers assisting us in achieving rapid scalability. A significant portion of this reinforcement comes from Microsoft, including engineers who have previously managed systems at comparable magnitudes."
"Our paramount priority is to provide a platform that is consistently available, trustworthy, and reliable for developers worldwide, and now, for the world’s AI agents as well," Daigle concluded.
Balancing Stability and Innovation: The Rationale Behind Continued Feature Development
A pertinent question arises: if the primary focus is on availability, why is GitHub continuing to launch new features, such as the new Copilot app showcased at its Build conference and other enhancements?
Daigle explained that not all product surfaces carry the same inherent risk. For instance, the Command Line Interface (CLI) and the new Copilot application operate independently of the core hosted GitHub platform. This allows for rapid iteration and development without impacting the critical systems undergoing repair. The backend work, he emphasized, is "focused on stability and resilience." Occasionally, the process of rebuilding the underlying architecture inadvertently facilitates the rollout of new features.
"If I release a new piece of functionality for the CLI, it does not possess the same stability and resilience characteristics as github.com," Daigle clarified.
While the availability of these new features would be moot if the underlying infrastructure were to fail, Daigle expressed optimism that this challenging phase in GitHub’s operational history will soon conclude. "We are hopeful that each subsequent month will bring incremental improvements, with the full urgency and dedication we can muster," he stated.
Broader Implications and the Future of Developer Infrastructure
The persistent outages at GitHub highlight a critical vulnerability in the modern software development ecosystem. As developers increasingly rely on AI-powered tools and automated workflows, the stability and reliability of foundational platforms like GitHub become paramount. The current situation underscores the immense challenge of scaling infrastructure to meet the exponential growth fueled by AI adoption.
The commitment to migrating to Azure and undertaking deep architectural overhauls signals a strategic shift towards greater resilience. The success of these initiatives will have far-reaching implications, not only for GitHub’s vast user base but also for the broader trajectory of developer tooling and cloud infrastructure. If GitHub can successfully navigate this period of intense demand and emerge with a more robust and scalable platform, it will set a precedent for how other critical developer services can adapt to the AI revolution.
The "all hands on deck" approach, bolstered by significant support from Microsoft, suggests a recognition of the systemic nature of the challenges. The focus on isolating critical services and optimizing underlying systems is a necessary evolution for platforms that have become indispensable to global software development. The eventual success of these efforts will be measured not by dramatic announcements, but by the quiet, consistent reliability that developers have come to expect, and now urgently need. The ongoing efforts to enhance stability, while simultaneously introducing new capabilities, represent a delicate balancing act, but one that is essential for maintaining developer trust and fostering continued innovation in the age of AI.
