Microsoft announced the Surface Laptop Ultra and Surface RTX Spark Dev Box, powered by Nvidia RTX Spark chips, with a focus on running AI models directly on the device instead of relying entirely on the cloud. In parallel, the company is introducing the Execution Containers feature for AI agents in Windows 11, with device sales set to begin on October 16.
Microsoft unveiled the Surface Laptop Ultra and the Surface RTX Spark Dev Box workstation, in a move that makes running AI models locally a core computer use case instead of restricting intensive processing to cloud services. The new devices rely on Nvidia RTX Spark chips and are aimed particularly at developers and users working with large models and demanding graphics workloads.
Devices Designed for Local AI
The Surface Laptop Ultra starts at $2,599, while the model equipped with a more powerful chip starts at around $3,700. Depending on the memory and storage options, the price may reach $5,900. The Surface RTX Spark Dev Box workstation, aimed at developers, starts at $6,000.
The workstation comes with tools including VS Code, GitHub Copilot CLI, Windows Subsystem for Linux, and PowerShell 7. Microsoft says the devices’ design combines the CPU, GPU, unified memory, and an improved cooling system to handle AI workloads directly on the device.
A Hybrid Approach Between the Device and the Cloud
Microsoft does not present the new devices as a complete replacement for the cloud. Instead, it is promoting a hybrid model that combines smaller models running locally with more powerful models hosted in the cloud. The company is also working on a new architecture that allows Copilot to use local and cloud models together.
The proposed use cases extend to content creation, video processing, and gaming, in addition to software development. The Surface Laptop Ultra and other devices based on RTX Spark are scheduled to go on sale beginning October 16. Microsoft is also offering a discount of up to $1,000 to users who trade in MacBook Pro devices, in a clear attempt to attract developers from Apple’s ecosystem to Windows.
Isolating AI Agent Tasks
Alongside the hardware, Microsoft announced the Execution Containers feature in Windows 11. The feature is intended to isolate operations performed by AI agents within a more secure execution environment. Satya Nadella said the feature will be available with the new Surface devices as well as to Windows 11 users.
Microsoft also intends to integrate Meta’s personal AI agent Muse into Windows. This indicates that the company views running agents not merely as a user-interface feature, but as a software workload requiring isolation and local computing resources.
What Is Changing in Practice?
The most significant shift is moving part of the AI burden from data centers to users’ computers. This could give developers greater ability to test models locally, but it comes with high prices. The source also does not specify performance details or the extent to which models and software are compatible with RTX Spark chips. Asus, Dell, HP, and Lenovo are also working on Windows devices built on the same platform; Dell has opened preorders for the XPS 16 Creator Edition at a price of $3,800.
With these products, Microsoft is seeking to broaden the definition of Copilot+ PC after its initial efforts faced a delay to the Recall feature because of security and privacy discussions. The limits of the protection provided by Execution Containers, and details regarding its availability to all Windows 11 users, still require further clarification.