Microsoft builds its own coding AI and cuts the OpenAI cord

Microsoft Build 2026 opened today in San Francisco and the headline was not Windows. Satya Nadella unveiled Project Polaris, a homegrown AI coding model that will replace GPT-4 Turbo as the default brain inside GitHub Copilot starting August 2026. Automatic migration for all Copilot subscribers, with a three-month fallback window for teams not ready to switch.

This is not a minor model swap. Polaris runs on Microsoft’s custom Maia 200 AI accelerators inside Azure, which the company says cuts per-inference latency and cost compared to routing through OpenAI. It was purpose-built for software development: code generation, multi-file refactoring, test writing, code review, documentation, dependency analysis. Microsoft called it “a peer programmer, not a pair programmer.” The framing matters. Copilot is no longer suggesting the next line. It gets assigned a bug or a feature and completes it independently.

The strategic calculation is blunt. The OpenAI-Copilot arrangement has always been awkward. Two companies with overlapping interests sharing a user base. OpenAI building Codex as a direct Copilot competitor. Microsoft competing with OpenAI on enterprise accounts. Polaris resolves the dependency. Microsoft now owns the model, the inference hardware, and the developer experience end-to-end.

That is not a theoretical future. It is already happening inside Microsoft. The company has been cancelling internal Claude Code licenses across its Experiences + Devices division, instructing engineers to move to GitHub Copilot CLI by June 30. Forbes, The Verge, and Times of India all confirmed the shift. Microsoft gave engineers Claude Code access in December. Five months later, the bill was too high and the strategy had changed. Uber reportedly burned through its entire 2026 AI tools budget on Claude Code and Cursor in four months. The economics of third-party AI coding tools are brutal at enterprise scale.

Project Polaris is Microsoft’s answer to that economics problem, and it doubles as a competitive weapon. Why pay Anthropic or OpenAI per token when you can run your own model on your own silicon? The Copilot SDK, in public preview since April, now ships with a reasoning layer built and operated entirely within Microsoft’s stack. Teams building on the SDK will be running Polaris whether they chose it or not.

Build 2026 had more than Polaris. The Windows Agent Framework 1.0 was open-sourced under MIT, giving developers a standard library for building agents across local Windows machines, Windows 365 Cloud PCs, and Azure Arc edge devices. Azure Agent Mesh was announced for federated multi-agent execution across cloud environments, including AWS Bedrock and Google Cloud AI Platform. Copilot Workspace graduated from beta to GA for GitHub Enterprise subscribers. Agent Mode became the default across Office 365 Copilot in Word, Excel, and PowerPoint. Foundry Local reached GA for on-device AI inference on Windows, macOS, and Linux.

The pattern across every announcement is the same. Microsoft is building the platform and owning the infrastructure. The MIT license on the Windows Agent Framework means developers can use any underlying model, but they build on Microsoft’s framework. Azure Agent Mesh integrates with AWS and Google Cloud, but governance flows through Entra ID and Purview. Copilot Workspace runs in GitHub, which Microsoft owns. Foundry Local uses the same API surface as Azure AI Foundry cloud, so model routing code works everywhere.

For developers using GitHub Copilot today: the GPT-4 fallback window closes at the November 2026 billing cycle. Evaluate your SDK integrations, test Polaris against your specific use cases, and set budget alerts now that usage-based billing went live June 1. The era of Microsoft relying on someone else’s model to power its developer tools is ending. The era of Microsoft owning the full stack has started.