Microsoft announced a new set of hardware and security tools on October 7, 2026, aimed at running more artificial-intelligence workloads directly on Windows computers. The company opened preorders for the Nvidia RTX Spark-powered Surface Laptop Ultra and made Microsoft Execution Containers generally available. The strategy is designed to move selected AI tasks from the cloud to local devices while giving organizations policy-based controls over what AI agents can access.
What the Surface Laptop Ultra includes
According to Microsoft’s official product announcement, the Surface Laptop Ultra is built around Nvidia’s RTX Spark platform. Top configurations combine a Grace CPU with as many as 20 cores, a Blackwell RTX GPU with up to 6,144 cores and as much as 128GB of unified memory. Microsoft says the system can deliver up to one petaflop of theoretical FP4 AI performance with sparsity and run models exceeding 120 billion parameters locally. Those figures are vendor specifications rather than results from independent benchmarks.
The laptop has a 15-inch touchscreen with peak HDR brightness of up to 2,000 nits, removable storage and support for as many as three external 4K displays. Ports include three USB-C connections, HDMI, USB-A, a full-size SD card reader and a headphone jack. Microsoft is also introducing a magnetic USB-C charging cable that attaches to one of the standard ports.
Surface Laptop Ultra starts at $2,599 in the United States. Preorders opened on October 7 and general availability is scheduled for October 16. The Verge reported that the highest configuration reaches $5,899.99. Prices, configurations and availability may differ by market, so the announced figures should be treated as part of the US launch rather than a universal price list.
Execution Containers for AI agents
Microsoft Execution Containers is a containment layer intended to limit AI agents running on a Windows PC. Microsoft says administrators can define which files, network resources and applications an agent may use, with Windows enforcing those rules on the device. The goal is to reduce the risk of an agent accessing data without authorization or carrying out actions beyond its assigned task.
Microsoft’s developer documentation organizes the security approach around three areas: containment to restrict an agent’s actions, identity to distinguish agent activity from human activity, and manageability to govern access and monitor behavior. The Execution Containers component is now generally available. Other capabilities, including Microsoft Entra-based separation of agent identities and the extension of Microsoft Agent 365 controls to local agents, are described as upcoming features rather than functions already available in full.
A hybrid local-and-cloud model
Microsoft calls the broader approach “hybrid intelligence.” A local computer can handle workloads that benefit from lower latency or keeping data on the device, while cloud resources remain available for tasks that require more capacity. The company also opened preorders for the Surface RTX Spark Dev Box at $5,999 in the United States, with shipments expected to begin in November.
Reuters described the announcement as an effort to turn Windows into a controlled platform for AI agents and to expand Nvidia’s position in a PC market long dominated by Intel and AMD. The report also identified rising memory costs and the price of high-end AI hardware as major obstacles to wider adoption. Microsoft and Nvidia must additionally demonstrate that local agents can be contained safely in real-world use.
The launch gives developers and creative professionals another option for running large models and coding tools without sending every task to a remote data center. However, actual performance, battery behavior, energy consumption and security effectiveness will depend on independent testing after the devices reach customers.
