Surface Laptop Ultra: Local AI hardware and Windows changes

Microsoft announced its new Surface Laptop Ultra, a high-end portable workstation starting at $2,599 that marks the debut of Nvidia’s Arm-based RTX Spark SoC. Shipping on October 16, the base model packs an 18-core CPU, 24GB of RAM, and 512GB of storage, while top-tier configurations reach up to $5,899 with 128GB of unified memory. Alongside the laptop, Microsoft revealed the Surface RTX Spark Dev Box, priced at $5,999, designed for developers running large local AI models.

How does Hybrid Intelligence change local AI execution?

According to Microsoft, the architecture determines whether a task runs locally on the device or via the cloud to optimize performance and token usage. For instance, tasks involving private files or high computational costs can execute on-device without sending documents over the internet. To keep these processes secure, Microsoft introduced Execution Containers (MXC), a sandboxing framework that assigns policy layers and isolates agent activities into specific backends like process or MicroVM containers.

What changes in the Windows 11 workflow?

The update brings natural language processing directly to the Windows search bar, allowing users to execute system-level actions like toggling dark mode or adjusting mic volume without digging through settings. Furthermore, Copilot is split into three tabs: Home for general LLM interaction, Code for no-code widget generation, and Autopilot for managing local AI agents that can handle tasks such as organizing folders and drafting emails.