Microsoft has reportedly turned to Amazon Web Services for additional GitHub compute capacity in June 2026, multiple sources close to the matter have confirmed. The move, which pairs two of the most intense cloud rivals, comes after explosive growth in AI-driven coding activity combined with repeated reliability problems on Microsoft’s own Azure platform left GitHub struggling to keep up with developer demand.
The decision, first detailed in a report shared with the Windows community, marks a rare public reliance on AWS infrastructure by Microsoft for a core service it owns. GitHub, acquired for $7.5 billion in 2018, is the world’s largest code-hosting platform, hosting over 200 million repositories and supporting tens of millions of developers. Its suite of AI tools—including GitHub Copilot and the Copilot Chat coding agent—has become central to modern software development, automatically generating code, reviewing pull requests, and even autonomously fixing bugs. But that AI revolution comes with a staggering compute appetite that has pushed Azure to its limits.
AI Coding Boom Overwhelms Azure
The trigger, sources say, was a perfect storm of surging AI usage and Azure reliability incidents. GitHub’s AI-powered features rely on large language models that demand continuous, high-performance GPU clusters for both model serving and fine-tuning. As Copilot’s user base swelled past millions of monthly active users and more enterprises adopted AI coding agents for automated development pipelines, the underlying infrastructure began to buckle. Azure data centers, particularly in regions serving North America and Europe, experienced a series of outages and degraded performance events in late 2025 and early 2026 that directly impacted GitHub Actions, Codespaces, and even basic Git operations.
These reliability problems weren’t just internal headaches. Developers reported sluggish response times from Copilot completions, failed CI/CD pipeline runs, and intermittent access to their repositories. Enterprise customers, who pay for GitHub Advanced Security and dedicated hosted runners, voiced frustration over missed SLAs and lost productivity. Microsoft’s traditional remedy—adding more Azure capacity—was constrained by global GPU shortages, supply chain delays for custom silicon, and intense internal competition for resources among Microsoft’s own AI services like Azure OpenAI and Microsoft 365 Copilot.
AWS to the Rescue: A Pragmatic Pivot
Faced with mounting reliability issues and a developer community quick to vent on social media and forums, Microsoft made the pragmatic choice to look beyond its own cloud. According to the report, Microsoft negotiated a large-scale capacity agreement with AWS to offload specific GitHub workloads. The contract, estimated to be worth hundreds of millions of dollars annually, provisions virtual machines, GPU instances, and likely AWS Lambda for serverless event processing—all to ensure GitHub’s AI features remain responsive and the core platform stays online during peak usage.
This isn’t the first time a Microsoft-owned property has leaned on AWS. After the LinkedIn acquisition in 2016, the professional network remained heavily dependent on AWS for years before a long, phased migration to Azure began. More recently, some Xbox cloud gaming services have quietly utilized AWS in regions where Azure presence is thin. But GitHub represents a far more visible and symbolic shift. The platform is deeply integrated with Microsoft’s developer ecosystem, from Visual Studio to Azure DevOps. Handing a piece of its compute to the competition underscores the severity of the capacity crunch.
The workloads being diverted to AWS are reportedly not the entirety of GitHub, but rather specific compute-intensive tasks tied to AI code generation and CI/CD execution. GitHub’s code storage, authentication, and core web serving remain on Azure. The arrangement also includes AWS’s proprietary chips—Trainium and Inferentia—which offer cost-effective AI acceleration that Microsoft’s own Azure AI infrastructure could not match at the required scale and speed.
What This Means for Azure and Microsoft’s Cloud Strategy
The optics are undeniably awkward. Microsoft has spent a decade and tens of billions of dollars building Azure into a credible AWS rival, boasting over 60 regions worldwide and serving 95% of Fortune 500 companies. Yet the company’s own flagship developer platform, the very forge where much of the world’s software is built, is now reliant on the competitor’s infrastructure. Satya Nadella has repeatedly championed Azure as the AI supercomputer of the future. The AWS deal suggests that future, for now, requires some help.
However, industry analysts see nuance. “This isn’t a failure of Azure as much as it is a testament to the insane pace of AI demand,” said one analyst. “No single cloud provider has infinite capacity, and even hyperscalers are bumping against physical constraints like power, real estate, and chip availability. Multi-cloud is the reality.” Indeed, enterprises have long embraced multi-cloud strategies to avoid vendor lock-in and ensure resilience. Microsoft itself offers Arc-enabled Azure services that run on AWS and Google Cloud. What makes this story unique is the public reversal: Microsoft is not just enabling customers to use AWS; it is doing so itself for a crown-jewel property.
There is precedent. Google, for example, uses Azure and AWS for certain YouTube infrastructure needs. Netflix relies heavily on AWS but also uses Google Cloud for some workloads. Even Amazon’s retail arm has been known to use Azure for experimentation. The hyperscaler walls are more porous than marketing materials suggest.
Developer Community Reacts with Irony and Concern
On Windows-focused forums and developer communities, the news has sparked a mix of irony, concern, and technical curiosity. “So the world’s largest code host is now running on the world’s most popular cloud,” one developer quipped, “which Microsoft has spent years saying is inferior to Azure.” Others expressed relief: “If this means my Copilot stop hallucinating and my Actions stop failing, I don’t care where it runs.”
Some power users worried about data sovereignty and security implications. GitHub retains immense amounts of source code, often proprietary and sensitive. Moving compute to AWS—even if the data itself stays encrypted at rest on Azure—raises questions about data in transit, multi-party access controls, and compliance with regulations like the EU’s GDPR. Microsoft and AWS have long-standing interoperability agreements for enterprise customers, but the scale of this integration will need rigorous auditing. GitHub assured in a preliminary statement that customer code would not be stored on AWS and that all data transfers would be encrypted and abide by existing data residency commitments.
The Hidden Costs of AI-Assisted Development
Underneath the irony is a deeper story about the hidden infrastructure costs of AI-assisted development. GitHub Copilot alone consumes immense resources: every keystroke-triggered suggestion, every code review summary, every agentic task execution. Microsoft has been operating Copilot’s backend on Azure GPU clusters that were already stretched thin by Bing Chat (now Copilot in Bing), Azure OpenAI Service customers, and Microsoft 365 Copilot. Adding GitHub’s growing AI workload created a zero-sum game where provisioning for one meant degrading another.
This capacity conflict is not unique to Microsoft. Google’s Gmail and Docs AI features compete for TPUs with Google Cloud customers. Amazon’s CodeWhisperer runs on—you guessed it—AWS. But AWS happens to have a larger capacity cushion in some regions and has invested earlier in custom AI silicon. By tapping Trainium and Inferentia, GitHub can achieve better price-performance for inference and fine-tuning, potentially reducing the cost per Copilot suggestion significantly.
For developers, the immediate benefit could be a more reliable experience. CI/CD pipelines that previously queued for minutes due to runner shortages may spin up faster on AWS instances. Codespaces environments, which let developers code in the browser with full compute backends, could become more responsive. And Copilot’s famously variable latency—sometimes near-instant, sometimes several seconds—might stabilize.
Will This Partnership Expand or Remain an Emergency Measure?
The critical question is whether the AWS tie-up is a temporary bridge or a long-term architectural shift. Microsoft has publicly committed to building out its AI infrastructure with massive capex—over $50 billion in the fiscal year ending June 2025, and more planned. New Azure regions with dedicated AI accelerators are coming online throughout 2026 and 2027. In theory, once that capacity arrives, GitHub could fully migrate back to Azure.
But the economics might argue otherwise. If AWS’s custom silicon consistently delivers lower cost-per-inference for Copilot’s specific model architectures, why move? Microsoft could find itself in the same situation as many of its enterprise customers: using Azure for what it does well and AWS for what it does better. That hybrid reality would be an ironic validation of the multi-cloud narrative that AWS has long promoted.
There’s also the politically sensitive dimension. Microsoft’s own Azure sales teams, who compete fiercely against AWS for every cloud deal, now have to explain to customers why GitHub—the developer experience they pitch as uniquely integrated with Azure—is actually powered by AWS in part. Competitors like GitLab, which runs on Google Cloud and touts its single-cloud consistency, may seize on the story to sway enterprise decisions.
The Bigger Picture: Cloud Rivals Collaborating in the AI Era
In the broader landscape, the AI boom is forcing unprecedented cooperation among cloud giants. Running frontier AI models at scale requires so much compute that no single provider can easily absorb demand. We’ve seen hints of this before: Microsoft and Oracle partnered to run Bing on Oracle Cloud Infrastructure during shortages; Google has offered its AI models on AWS Bedrock. What’s different here is the optics: GitHub is a developer-loved brand, and developers are a notoriously opinionated audience that remembers promises.
Microsoft’s long-term solution likely involves three tracks: doubling down on its own custom silicon (the rumored Athena AI chips), continuing strategic use of AWS for peak shaving, and aggressively expanding Azure’s AI capacity. In the meantime, the GitHub on AWS story will serve as a case study for years on the industry’s capacity limits and the blurring lines between competitors.
For the millions of developers pushing code to GitHub daily, the bottom line is simpler: if it works, they’ll keep using it. The AWS deal is, above all, a reliability play. As one commenter on a Windows insider forum put it, “I don’t care if my code runs on a toaster, as long as it doesn’t drop my commits.” That pragmatism, shared by coders everywhere, may be the ultimate justification for this strange-bedfellows alliance.
The coming months will test whether the added capacity truly stabilizes GitHub and how Microsoft juggles the competing demands of its AI ambitions. But one thing is certain: the server racks spinning up GitHub’s next Copilot suggestion may now be painted orange instead of blue.