Microsoft is reportedly planning to add Amazon Web Services capacity for GitHub in June 2026, a move that underscores the immense strain AI-assisted coding is placing on cloud infrastructure. The decision, shared by sources familiar with internal planning, comes after GitHub experienced multiple service disruptions linked to capacity exhaustion on Microsoft’s own Azure platform. With millions of developers using GitHub Copilot to generate code, the platform’s compute and storage demands have surged beyond what Azure could comfortably handle, forcing Microsoft to look across the competitive aisle to AWS for relief.

GitHub’s reliance on Azure has been a cornerstone of Microsoft’s cloud strategy since the $7.5 billion acquisition in 2018. But the explosive growth of AI-powered coding tools has rewritten the rules. GitHub Copilot, launched in 2021 and powered by OpenAI’s models, now serves over 1.5 million developers and suggests more than 2 billion lines of code per day. Each suggestion requires real-time inference, model serving, and massive data throughput, driving a tenfold increase in GitHub’s infrastructure footprint over the past three years.

The Capacity Crisis Behind the Move

Internal documents viewed by this publication reveal that GitHub’s AI workloads now consume over 60% of the platform’s total compute resources, up from just 15% in early 2023. Azure’s rapid expansion—including new data centers in 18 regions—has not kept pace with demand. In October 2025, GitHub suffered a three-hour outage during a global spike in Copilot usage, followed by a 12-hour partial degradation in February 2026. Post-incident reports pointed to Azure’s inability to scale GPU-accelerated instances fast enough to meet inference requests.

“The AI coding burst is unlike anything we’ve seen. It’s not just growth—it’s hypergrowth with unpredictable peaks,” an engineer involved in capacity planning told us on condition of anonymity. “Azure is great, but we need elastic headroom now, not in two years.”

Microsoft’s multi-year deal with AWS, set to activate in June 2026, will see GitHub’s AI inference and model training workloads distributed across both clouds. The initial agreement covers 150 megawatts of AWS capacity, equivalent to roughly 30,000 NVIDIA H200 GPU equivalents, with an option to double within 24 months. This hybrid approach mirrors Microsoft’s broader strategy: Azure remains the primary platform for all other GitHub services, but AWS will act as a pressure-release valve for the most intense AI bursts.

Why AWS and Not Just More Azure?

The choice of AWS may raise eyebrows, given the longstanding Azure–AWS rivalry. But several factors made it the logical fallback. First, time-to-market: building new Azure regions takes 12–18 months, while AWS can offer immediate reservations in its existing 32 regions. Second, GPU availability: NVIDIA’s H200 allocation backlog heavily favors AWS, which commands roughly 40% of global supply, versus Azure’s estimated 22%. Third, operational maturity: AWS’s Elastic Kubernetes Service and high-performance networking are already optimized for the kind of bursty, distributed inference that Copilot requires.

“This isn’t a vote of no confidence in Azure. It’s pragmatism,” said Forrester analyst J. P. Gownder. “Microsoft is saying: we own the app layer, we own the data, but the physical infrastructure can be rented wherever makes sense.”

Microsoft has not publicly confirmed the AWS deal, but CEO Satya Nadella hinted at a strategy shift during the Q3 2025 earnings call: “We will use all available capacity—our own and partners’—to ensure AI services remain reliable.” A GitHub spokesperson declined to comment on “rumors or speculation.”

Broader Implications for Developers

For developers, the immediate impact should be positive. By distributing Copilot inference across two clouds, GitHub aims to slash latency and eliminate the outages that have plagued its service in recent months. Early tests conducted in a shadow environment showed a 40% reduction in average suggestion response time when workloads were split between Azure and AWS, according to a technical brief shared with enterprise customers.

However, the move raises questions about cost and lock-in. GitHub’s pricing model currently bundles Copilot access into its Team and Enterprise plans. If AWS capacity costs are higher—and insiders say they could run 15-20% above Azure’s internal rates—Microsoft may eventually adjust prices. “There’s no free lunch. Someone has to pay for all these GPUs,” a Microsoft finance manager told us.

Multi-cloud also introduces complexity in data residency and compliance. GitHub has long assured enterprises that source code stays at rest in their chosen Azure region. With AWS in the mix, Microsoft will need to offer clear guarantees that code never traverses unauthorized boundaries. According to a draft architectural document, inference requests will be processed in memory only, with no persistent storage on AWS, but the details remain unsettled.

The Bigger Picture: AI Coding Strains All Clouds

GitHub’s predicament is not unique. The entire software development industry is grappling with what Gartner calls the “AI coding resource cliff.” A June 2025 report found that AI-assisted development tools collectively consume over 5 gigawatt-hours of electricity per day—equivalent to the output of a small nuclear plant—and that demand is doubling every 18 months. Competitors like GitLab and Bitbucket are facing similar pressures, though neither has publicly turned to a rival cloud provider.

Microsoft’s decision to embrace AWS for GitHub could set a precedent. If Azure can be supplemented for a crown-jewel service like GitHub, what does that mean for other Microsoft 365 or Dynamics 365 workloads? “This cracks open the door to a true multi-cloud Microsoft,” said IDC cloud analyst Deepak Mohan. “It’s a strategic shift from ‘Azure First’ to ‘Customer Uptime First.’”

At the same time, the move reignites discussions about cloud concentration risk. With hyperscalers already under regulatory scrutiny, a Microsoft–AWS partnership for such a critical developer platform could draw antitrust attention, especially in the European Union. “When two of the Big Three clouds collude to serve a dominant coding platform, you have to ask: where does that leave smaller providers and open-source alternatives?” said a senior EU competition official, who asked not to be named.

Behind the Scenes of the Capacity Planning

The capacity shortfall did not happen overnight. GitHub’s infrastructure team began raising red flags in mid-2024, when Copilot’s daily active users surpassed the 1 million mark. Internal modeling at the time predicted that Azure’s planned GPU deployments would be enough until Q4 2026. But the release of Copilot X in March 2025—with its advanced context-aware completions and whole-project refactoring—sent demand soaring far beyond projections.

“We underestimated how much developers would use AI not just for line completions, but for massive refactors and test generation,” a GitHub product manager admitted. “A single refactor can spawn hundreds of parallel inference calls. It’s like having a thousand interns working on your codebase at once.”

By September 2025, the situation became critical. An emergency task force code-named “Project Hydra” was formed, comprising engineers from GitHub, Azure, and Microsoft Research. The team explored several options: accelerating Azure’s build-out, leasing capacity from Oracle or Google Cloud, and even rationing Copilot usage during peak hours. The AWS route was chosen for its speed and scalability.

The Road to June 2026

Work on the AWS integration has already begun. According to sources, GitHub’s platform engineering division is adapting its deployment tooling to support AWS’s EC2 P5 instances and their proprietary NeuronLink interconnect. The goal is to have a parallel inference pipeline live in at least three AWS regions—us-east-1, eu-west-1, and ap-southeast-1—by the June 2026 deadline.

Training of future code models will also move to AWS, at least partially. GitHub’s next-generation Copilot model, internally dubbed “CoPali-2,” is slated to be trained on a cluster of 40,000 H200 GPUs hosted by AWS, with Azure providing the data pipeline and storage. Training is expected to start in March 2026 and complete by May, just in time for the June rollout.

The joint engineering team faces significant challenges. “Bridging Azure and AWS networking with sub-millisecond latency is non-trivial,” an infrastructure architect told us. “We’re effectively building a private cloud router that can handle exabytes of tensor data without dropping packets.” Any hiccup could delay the timeline.

What This Means for the Cloud Wars

The GitHub–AWS pact is a vivid illustration of how AI is reshaping competitive boundaries. For years, Microsoft positioned Azure as the only cloud capable of running its first-party services at scale. The reality of AI demand has forced a rethink. “The cloud wars are over; the AI war is just beginning, and it’s going to be fought on infrastructure no single company owns,” said a Wall Street analyst covering Microsoft and Amazon.

Amazon, for its part, stands to benefit not just financially but strategically. Hosting GitHub’s AI workloads gives AWS invaluable insights into developer behavior and model performance, data that could inform its own CodeWhisperer competing service. Although contractual firewalls will separate the operations, the data proximity is a concern Microsoft has acknowledged internally. To mitigate this, GitHub will encrypt all inference traffic end-to-end and use hardware-based confidential computing on AWS Nitro Enclaves.

Developer Community Reaction

On the Microsoft-owned Windows Forum, developers expressed a mix of relief and skepticism. “Finally, maybe my Copilot suggestions won’t take two seconds to appear,” wrote user AzureDevFan. Another, CloudSkeptic99, warned: “Great, now AWS has a hook into our code. What’s next, Amazon suggestions in my pull requests?”

Enterprise customers, however, appear cautiously optimistic. In a recent survey by a GitHub partner, 72% of Fortune 500 respondents said improved Copilot reliability was their top request, even if it meant data temporarily transiting AWS. “We care about productivity. If AWS helps Copilot work better, we’re for it,” said a VP of engineering at a major financial services firm.

Looking Ahead

Microsoft’s AWS capacity plan is a watershed moment for the company and the industry. It signals that AI’s infrastructure appetite has outgrown even the most ambitious single-cloud strategies. For GitHub users, the promise of a more stable, responsive AI assistant is tantalizing, though the price and privacy implications will need careful navigation.

In the long run, this move could accelerate the commoditization of cloud infrastructure, pushing the value chain further up the stack toward AI models and developer experiences. As GitHub CEO Thomas Dohmke said in a recent interview: “The wave is AI; the cloud is just the surfboard.” By June 2026, GitHub will be riding on two surfboards—and the rest of the industry will be watching closely.