Updated

Updated · ZDNet · Jun 15

Microsoft Unveils WSL 3 Beta for Windows 11, Cutting AI Overhead With Direct GPU and NPU Access

Updated

Updated · ZDNet · Jun 15

Microsoft Unveils WSL 3 Beta for Windows 11, Cutting AI Overhead With Direct GPU and NPU Access

Q: With WSL 3, is Microsoft admitting Linux is the true future for AI, even on its own PCs?

No. WSL 3 is less an admission that Linux will replace Windows for AI than a recognition that the most important AI developer tools and frameworks are still largely Linux-first. Microsoft’s move is pragmatic. PyTorch, TensorFlow, Ollama, llama.cpp and many CUDA-centered workflows run best in Linux environments, and WSL 2 imposed a noticeable virtualization penalty, especially around GPU, NPU and file-system performance. WSL 3’s new paravirtualized design is meant to remove much of that overhead, giving Linux workloads near-native access to Windows hardware. That is a technical acknowledgment that developers want Linux compatibility without giving up Windows PCs. The broader strategy is to keep Windows relevant as an AI development platform. Microsoft is pairing WSL 3 with Copilot+ PCs, Windows AI APIs, local AI hardware, WSL containers and other tooling so developers can run Linux-based stacks on top of Windows rather than leave Windows entirely. In that sense, Linux is being treated as a core layer in the Windows developer story, not as a rival operating system that has won outright. The stronger takeaway is that Microsoft now sees interoperability as essential. In the cloud, Linux already underpins much of Azure and major Microsoft services. On PCs, WSL 3 extends that reality to local AI work: the future Microsoft is betting on is Windows as the host platform, with Linux as a critical runtime for serious AI development.

2 articles · Updated · ZDNet · Jun 15

WSL 3 entered beta at Build 2026 as Microsoft’s next Linux-on-Windows layer, aimed at speeding Linux AI workloads on Windows 11 rather than replacing the existing WSL shell experience.
The gain comes from a new paravirtualized VM design that gives Linux processes more direct access to GPUs and NPUs, reducing the translation and context-switching overhead that limited WSL 2.
Microsoft said Linux containers will run directly on Windows with no extra configuration, and frameworks such as PyTorch and TensorFlow should perform closer to native Linux—though bare-metal Linux remains faster.
Early support targets Copilot+ PCs and systems using Qualcomm Snapdragon X Elite, Intel Meteor Lake and Lunar Lake chips; AMD support will not be available at first.
WSL 3 will roll out through Windows Insider and pre-release WSL packages before broader Windows 11 adoption, with Microsoft positioning it as a core tool for developers who need Linux AI stacks without leaving Windows.