Claude Code, Anthropic’s agentic coding tool, spread rapidly across Microsoft’s engineering organization in early 2026, according to reporting by The Verge’s Notepad newsletter. While developers have spent months comparing Claude Code, Anysphere’s Cursor, and Microsoft’s GitHub Copilot without crowning a single winner, Claude Code’s ease of use — for developers and non-developers alike — reportedly pushed it to the front inside the company that makes its biggest competitor.
How Claude Code spread inside Microsoft
Microsoft first adopted Anthropic’s Claude Sonnet 4 model inside its developer division in mid-2025, then brought Anthropic models to paid GitHub Copilot users several months later. The internal push went further: Microsoft’s CoreAI team — the AI engineering group led by former Meta engineering chief Jay Parikh — spent months testing Claude Code, and the Experiences + Devices division, responsible for Windows, Microsoft 365, Outlook, Teams, Bing, Edge, and Surface, was subsequently asked to install it.
Notably, even employees with no coding experience were encouraged to experiment, letting designers and product managers prototype ideas. Microsoft also approved Claude Code for use across the code and repositories of its business and industry Copilot teams, and engineers were reportedly expected to use both Claude Code and GitHub Copilot and give comparative feedback.
An awkward position for GitHub Copilot
Microsoft sells GitHub Copilot as the AI coding tool of choice for its customers, and the company has previously said that 91 percent of its engineering teams use it. Widespread internal enthusiasm for a rival tool therefore carried obvious tension. Microsoft communications chief Frank Shaw framed the pilots as routine, saying companies regularly test competing products to understand the market, and stressing that OpenAI remained Microsoft’s primary frontier-model partner.
At the same time, the commercial relationship with Anthropic deepened. A November 2025 agreement brought Claude Sonnet 4.5, Claude Opus 4.1, and Claude Haiku 4.5 to Microsoft Foundry customers, alongside a commitment by Anthropic to purchase $30 billion of Azure compute capacity, and Microsoft began counting Anthropic model sales toward Azure sales quotas. Microsoft has also selectively used Anthropic models inside Microsoft 365 apps and Copilot features where they proved more capable than OpenAI counterparts.
What happened next: costs forced a reversal
The experiment did not stay unconstrained for long. By mid-2026, multiple outlets reported that Microsoft was cancelling most internal Claude Code licenses in the Experiences + Devices division effective June 30 — the end of its fiscal year — and steering engineers toward GitHub Copilot CLI instead. Reporting by Windows Central and TechRadar pointed to cost as the main driver: Claude Code’s usage-based token billing meant spending scaled with popularity, while GitHub Copilot’s flat per-seat pricing is predictable. The tool’s internal success, in other words, made it expensive — and it competed directly with the product Microsoft wants customers to buy.
Why this episode matters
The arc is a useful case study in enterprise AI adoption. It shows how quickly agentic coding tools can move from pilot to daily workflow when they lower the barrier for non-technical staff; how usage-based pricing can collide with annual budgeting once adoption takes off; and how strategic considerations — selling one’s own product — can override engineer preference. It also illustrates the pressure these tools place on traditional development roles, as more code generation shifts to autonomous agents. Broader governance questions raised by that shift are covered in this post on governance in the agentic AI boom, and the cost dynamics connect to this tutorial on agents operating under token budgets.
Limitations and what to watch
Much of this account rests on reporting from unnamed sources and internal communications reviewed by journalists, and Microsoft has publicly characterized the adoption as routine product testing rather than a strategic shift. Figures such as the share of engineering teams using Copilot come from Microsoft itself. What to watch: whether Microsoft’s Copilot CLI closes the capability gap engineers reportedly perceived, whether Anthropic adjusts enterprise pricing in response to cost-driven churn, and whether other large companies follow the same adopt-then-retrench pattern as AI tool budgets mature.
