Microsoft has always been at the forefront of generative AI, undoubtedly driven by its massive investment in ChatGPT manufacturer OpenAI. The company has launched AI tools for consumers and developers, including the recently announced Windows AI Studio. However, developers hoping to try out Microsoft's latest AI tool may be surprised to find that it does not run on Windows. Windows AI Studio requires Linux.

Windows AI Studio is another attempt by Microsoft to encourage local Windows app development. AI Studio provides developers with a toolkit to integrate generative AI into their applications. It includes templates and step-by-step guides to implement advanced language models like Microsoft's Phi and Meta's Llama 2.

It may seem strange that Microsoft specifies Linux as a requirement for the development tool, but Linux is not actually in competition with Windows. Linux is common in development and web hosting environments, but the era of Linux desktop may never come. However, you can run Linux within Windows, which is exactly what Microsoft suggests for Windows AI Studio.

To use this tool, your system needs to run Ubuntu 18.04 or higher, and luckily, Windows Subsystem for Linux (WSL) can meet this requirement. The new AI tool is distributed as a Visual Studio extension, so you need to install VS Code first. Microsoft designed WSL to enable them to interact with each other. VS Code can run on Windows while using WSL as the data source.

This tool has another important limitation, but it is not surprising. Windows AI Studio requires an Nvidia GPU. Although theoretically, all video cards are suitable for machine learning workloads because of their highly parallel architecture, Nvidia's significant investment in AI makes its cards the preferred AI development platform.

Windows AI Studio is still very new, and we expect to see stronger hardware support in the future. Microsoft has indicated that the tool will eventually optimize models for various Windows GPUs and NPUs, and we will see more with the arrival of new CPUs and SoCs with desktop machine learning hardware. Nevertheless, you may still need Linux for development.