A platform overwhelmed by machine-made code

The root cause is a staggering demand shock from agentic coding. GitHub wasn't built for software that writes, commits, and deploys itself at hyper-scale.

• AI agent pull requests exploded 14x. Agent-opened pull requests catapulted from 4 million in September 2025 to more than 17 million by March 2026 . The volume of commits has completely rewritten GitHub's operational reality: the platform now handles about 275 million commits per week, a figure that exceeds the 1 billion commits it processed in all of 2025 .
• Reliability collapsed. Between May 2025 and April 2026, GitHub suffered 257 incidents, 48 of which were classified as severe—an average of one major outage per week . The strain intensified in 2026. Nine service incidents were logged in May alone, and by June, GitHub's availability had dropped to roughly 88.4%, a dramatic fall from its 99.9% service level agreement target .
• Leadership acknowledged the breaking point. During Microsoft's April 2026 earnings call, CEO Satya Nadella stated GitHub was "experiencing unprecedented growth, driven by the proliferation of agentic coding" and that the company was "diligently working to scale and meet this demand" . Earlier, GitHub's CTO Vlad Fedorov had published a rare public apology, admitting the platform badly underestimated its capacity needs and that April uptime had slipped below 85% .

The immediate trigger was clear: GitHub's infrastructure simply could not absorb the load. The platform's infrastructure team had planned for a 10x capacity expansion in October 2025 but realized by February 2026 that they needed a 30x increase .

Turning to a rival: the AWS stopgap

With weekly outages disrupting millions of developers and AI agents alike, Microsoft made an operational decision, not a strategic one. It bought spot capacity from AWS to stabilize the platform .

A Microsoft spokesperson confirmed GitHub is tapping multiple cloud providers but declined to comment specifically on Amazon's involvement, saying only: "The incredible spike in agentic development that began late last year has tested our infrastructure" . The AWS capacity is described as a temporary measure to relieve immediate strain while the longer-term migration to Azure continues .

The irony is heavy. Microsoft acquired GitHub for $7.5 billion in 2018 with the explicit goal of weaving it into Azure and competing more directly with AWS in the cloud wars . Now, a product of the AI boom—much of it powered by GitHub's own Copilot—has forced Microsoft's hand to pay its arch-rival for servers.

The stalled Azure migration: a plan outpaced by reality

The AWS deal is a billboard-sized signal that GitHub's Azure migration is behind schedule and outmatched by demand. After the 2018 acquisition, Microsoft set a course to migrate all of GitHub's infrastructure from its legacy data centers to Azure . That migration is a multi-year effort:

• GitHub's CI/CD service (Actions) and Copilot models were migrated to Azure in 2024 .
• Pages and Packages followed in 2025 .
• The core platform and databases, however, only began migration in October 2025 . By February 2026, only 12.5% of GitHub's traffic was routing through Azure, with a stated target of 50% by July 2026 .

The original plan was to have GitHub completely on Azure by 2027 . But the load curve from AI coding has moved faster than the migration timeline . GitHub's CTO acknowledged the platform remains tethered to older data centers, limiting its ability to scale rapidly during a traffic explosion .

Adding capacity from AWS suggests Azure either lacked the available compute in the right geographic regions or simply could not provision it fast enough to stop the bleeding . For a company that sees Azure as its future, this is a profound operational concession.

Operational responses: more than just a Band-Aid

Beyond the headline-grabbing AWS rental, GitHub and Microsoft are implementing several operational changes to fix the underlying fragility:

1. A forced pivot to multi-cloud. GitHub's infrastructure is now, de facto, multi-cloud. Microsoft is using AWS capacity as an emergency surge layer while the Azure migration continues . This is a tactical, not idealistic, architecture.
2. Accelerating migration tooling. On the Azure DevOps side, Microsoft recently launched a preview of Enterprise Live Migrations (ELM), a tool designed to migrate multiple repositories with minimal cutover downtime, signaling that migration infrastructure is being upgraded under pressure .
3. Existential internal reprioritization. Earlier reports indicated GitHub temporarily halted most new feature development to focus entirely on its "existential" core-infrastructure migration to Azure, driven by capacity limits in its Virginia data centers .

The takeaway is not that Microsoft is abandoning Azure for GitHub—it isn't. The takeaway is that AI coding agents have permanently changed the scale of developer infrastructure. GitHub's load was designed for humans typing into keyboards. It is now being re-architected on the fly for a world where software writes itself, and even a trillion-dollar company like Microsoft can be caught flat-footed, forced to rely on its biggest competitor just to keep the lights on.

A platform overwhelmed by machine-made code

The root cause is a staggering demand shock from agentic coding. GitHub wasn't built for software that writes, commits, and deploys itself at hyper-scale.

• AI agent pull requests exploded 14x. Agent-opened pull requests catapulted from 4 million in September 2025 to more than 17 million by March 2026 . The volume of commits has completely rewritten GitHub's operational reality: the platform now handles about 275 million commits per week, a figure that exceeds the 1 billion commits it processed in all of 2025 .
• Reliability collapsed. Between May 2025 and April 2026, GitHub suffered 257 incidents, 48 of which were classified as severe—an average of one major outage per week . The strain intensified in 2026. Nine service incidents were logged in May alone, and by June, GitHub's availability had dropped to roughly 88.4%, a dramatic fall from its 99.9% service level agreement target .
• Leadership acknowledged the breaking point. During Microsoft's April 2026 earnings call, CEO Satya Nadella stated GitHub was "experiencing unprecedented growth, driven by the proliferation of agentic coding" and that the company was "diligently working to scale and meet this demand" . Earlier, GitHub's CTO Vlad Fedorov had published a rare public apology, admitting the platform badly underestimated its capacity needs and that April uptime had slipped below 85% .

The immediate trigger was clear: GitHub's infrastructure simply could not absorb the load. The platform's infrastructure team had planned for a 10x capacity expansion in October 2025 but realized by February 2026 that they needed a 30x increase .

Turning to a rival: the AWS stopgap

With weekly outages disrupting millions of developers and AI agents alike, Microsoft made an operational decision, not a strategic one. It bought spot capacity from AWS to stabilize the platform .

A Microsoft spokesperson confirmed GitHub is tapping multiple cloud providers but declined to comment specifically on Amazon's involvement, saying only: "The incredible spike in agentic development that began late last year has tested our infrastructure" . The AWS capacity is described as a temporary measure to relieve immediate strain while the longer-term migration to Azure continues .

The irony is heavy. Microsoft acquired GitHub for $7.5 billion in 2018 with the explicit goal of weaving it into Azure and competing more directly with AWS in the cloud wars . Now, a product of the AI boom—much of it powered by GitHub's own Copilot—has forced Microsoft's hand to pay its arch-rival for servers.

The stalled Azure migration: a plan outpaced by reality

The AWS deal is a billboard-sized signal that GitHub's Azure migration is behind schedule and outmatched by demand. After the 2018 acquisition, Microsoft set a course to migrate all of GitHub's infrastructure from its legacy data centers to Azure . That migration is a multi-year effort:

• GitHub's CI/CD service (Actions) and Copilot models were migrated to Azure in 2024 .
• Pages and Packages followed in 2025 .
• The core platform and databases, however, only began migration in October 2025 . By February 2026, only 12.5% of GitHub's traffic was routing through Azure, with a stated target of 50% by July 2026 .

The original plan was to have GitHub completely on Azure by 2027 . But the load curve from AI coding has moved faster than the migration timeline . GitHub's CTO acknowledged the platform remains tethered to older data centers, limiting its ability to scale rapidly during a traffic explosion .

Adding capacity from AWS suggests Azure either lacked the available compute in the right geographic regions or simply could not provision it fast enough to stop the bleeding . For a company that sees Azure as its future, this is a profound operational concession.

Operational responses: more than just a Band-Aid

Beyond the headline-grabbing AWS rental, GitHub and Microsoft are implementing several operational changes to fix the underlying fragility:

1. A forced pivot to multi-cloud. GitHub's infrastructure is now, de facto, multi-cloud. Microsoft is using AWS capacity as an emergency surge layer while the Azure migration continues . This is a tactical, not idealistic, architecture.
2. Accelerating migration tooling. On the Azure DevOps side, Microsoft recently launched a preview of Enterprise Live Migrations (ELM), a tool designed to migrate multiple repositories with minimal cutover downtime, signaling that migration infrastructure is being upgraded under pressure .
3. Existential internal reprioritization. Earlier reports indicated GitHub temporarily halted most new feature development to focus entirely on its "existential" core-infrastructure migration to Azure, driven by capacity limits in its Virginia data centers .

The takeaway is not that Microsoft is abandoning Azure for GitHub—it isn't. The takeaway is that AI coding agents have permanently changed the scale of developer infrastructure. GitHub's load was designed for humans typing into keyboards. It is now being re-architected on the fly for a world where software writes itself, and even a trillion-dollar company like Microsoft can be caught flat-footed, forced to rely on its biggest competitor just to keep the lights on.