It can take organizations several hours to set up brand-new PCs for employees, and programmers in particular. With Project Zenith, Microsoft hopes to drastically reduce that time. However, the configuration and suite of pre-installed tools are based on a specific profile that is heavily AI-focused.
Zenith makes its debut on AMD Ryzen AI Halo systems, compact PCs with a remarkable amount of AI capabilities for their size. 64 GB of unified memory and at least 250 GB/s of memory bandwidth, the minimum requirement for Zenith, make it possible to run AI locally at an acceptable speed. Microsoft states that users don’t always want to wait for this on the standard Windows configuration.
With Project Zenith, a variant of Windows 11 is coming that, according to the company, is immediately ready for coding, without the usual hours of upfront configuration work. Logan Iyer, Corporate Vice President of Windows Platform + Developer, describes Zenith as a “ready-to-code Windows experience.”
Pre-configured and streamlined
While the changes are significant compared to a “standard” Windows installation, they’re also simple. They’re mostly very obvious tweaks that can be set up individually with just a few mouse clicks. On Zenith machines, for example, Windows Terminal and Visual Studio Code are pinned to the taskbar by default. File Explorer displays file extensions, hidden files, and the full path, with support for long path names. Hints in the Start menu and account notifications are disabled. Command Palette, on the other hand, is enabled.
The inclusion of the Windows Subsystem for Linux (WSL) is a bit more in-depth. Microsoft previously brought WSL to the Microsoft Store as the default distribution channel and added Fedora as an officially supported distribution. Last year, the source code was opened up. At Build 2026, WSL containers were introduced, allowing Linux containers to run directly from within Windows. WSL is a way for Microsoft to retain developers as Windows users while also providing them with the Linux distributions and functionality they desire.
Local models vs. cloud costs
As mentioned, the focus is on AI. Microsoft wants developers to run smaller coding models locally and without limits, so fewer cloud tokens are needed. Major challenges remain with frontier models. For developer purposes, Visual Studio Code, GitHub Copilot, PowerShell, Git, Python, NVM, and Node, among others, are also installed.
AMD is supplying the first platform. The manufacturer claims that Ryzen AI Halo, with 128 GB of unified memory, can handle models with up to 200 billion parameters, though many real-world tests put the limit at around 70 billion; this refers to full-scale models (i.e., not compressed via quantization, which results in poorer outputs) with sufficient KV cache, or short-term memory, to function properly. Developers who want to work locally currently choose mainly between NVIDIA’s DGX Spark with the CUDA stack and Apple’s Mac Studio with greater memory bandwidth. These represent two extremes: dev kits on one end and virtually consumer-grade machines on the other. A Windows system like Ryzen AI Halo fits somewhere in between.
Zenith devices will also come with the agent-based security that Microsoft demonstrated at Build 2026 from day one, including Microsoft Execution Containers. More devices from OEM and chip partners will follow in the coming months.