Microsoft is doubling down on AI development for Windows with a major update aimed at empowering app builders. At its Build 2025 conference, the tech giant unveiled Windows AI Foundry, a newly branded and expanded platform that promises to make local AI development on Windows much easier and more accessible.
The initiative is a significant upgrade to what was formerly known as Windows Copilot Runtime, introduced last May. Now reimagined as a unified toolkit, Windows AI Foundry is designed to help developers fine-tune, optimize, and deploy AI models directly within Windows environments.
Why is this important? Microsoft sees AI not just as a feature—but as a foundational layer for the future of computing. Despite the steep costs of training large-scale AI models, the company’s strategy is starting to pay off. CEO Satya Nadella recently revealed that Microsoft is generating $13 billion in annualized revenue from AI—thanks in large part to its partnership with OpenAI.
That said, Microsoft isn’t aiming to make money directly from third-party AI apps built on Windows. Instead, the goal is to encourage a thriving ecosystem of AI-powered tools running natively on its platform. The easier it becomes to develop these tools for Windows, the more developers are likely to stay in Microsoft’s orbit—and not drift to rivals.
Windows AI Foundry addresses many of the pain points developers face today. It automatically detects a device’s hardware setup and pulls in the correct software components needed to run AI models efficiently. It even keeps those components up to date as hardware evolves, making life easier for developers who want to focus on building, not maintaining.
A standout addition to the platform is Foundry Local, a new module designed to bring AI capabilities directly to user devices. Compatible with both Windows and macOS, Foundry Local lets developers run models locally using ONNX Runtime, a popular open-source framework known for its cross-platform AI performance.
Foundry Local also includes a command-line interface that allows developers to quickly browse available models, test them, and run them on local machines using simple prompts like “Foundry model list” or “Foundry model run.” It even detects available hardware—CPU, GPU, or NPU—and displays compatible models that can be deployed with ease.
What’s more, developers can use the Foundry Local SDK to integrate these features into their apps, opening up new possibilities for responsive, privacy-conscious, and offline-capable AI experiences.
With Windows AI Foundry, Microsoft is signaling a clear message: AI development should be simple, fast, and deeply embedded in the operating system itself. And by removing friction from the process, the company is betting it can win the hearts—and workloads—of developers everywhere.
