October 8, 2026
BlogMicrosoft Makes Its Case (Again) for Running AI Workloads on Windows PCs
Microsoft is trying to make the idea of running AI locally on Windows PCs appeal to users beyond just developers. The company is hoping more people might be swayed by potential cost and convenience benefits of local processing of apps and agents.Â
As part of this attempt, Microsoft is building more AI-focused features into Windows 11 to make the platform better at running and managing agents. And it is pushing the new Copilot “super app,” announced in late Sept., as a vehicle for making computing on Windows 11 more personal and productive.
Company officials showcased these capabilities during a Windows and Surface news event on Oct. 7.
One of the agent-focused Windows 11 capabilities highlighted during the event, Microsoft Execution Containers (MXC), is generally available to commercial customers as of this week.
Microsoft first took the wraps off the MXC software development kit (SDK) at its Build developers conference in early June, announcing an early preview of the technology. At that time, Microsoft officials said MXC provided “an abstraction layer across isolation primitives, so developers do not have to manage low-level isolation details.”
MXC enables developers to set the resources, like files and network destinations an agent has permission to use. MXC selects the appropriate container to enforce those policies at runtime, as Microsoft explains. With agents, MXC can be used to isolate model-generated output, plug-ins, tools, harnesses, and/or entire agents.
MXC works with Windows 11, macOS, Linux, and MicroVMs, but offers the most options on Windows, where both process and session containers are supported.
Microsoft officials have said they plan to integrate MXC with Intune to support governing agent workloads on Windows devices. They also touted the coming Entra and Agent 365 integration with Windows 11 that will help distinguish agents from human users — a step towards Microsoft’s stated goal of licensing many agents similar to the way they license people by providing them with their own Entra IDs, email addresses, Teams accounts, and more.
All Hail ‘Hybrid Intelligence’Â
Security wasn’t the only touchstone at the Oct. 7 event. Microsoft officials also showcased efficiency and cost.
New intelligent model-routing capabilities built on GitHub HydraFusion,are also coming to the GitHub Copilot app, GitHub Copilot CLI and Visual Studio Code in the form of experimental previews later in October. This routing capability will figure out whether a task is best suited to local, cloud, or hybrid computing. Microsoft’s new buzzword for AI processing that happens across local and cloud is “hybrid intelligence.”
In addition to Microsoft’s own MAI Code 1.1 Flash model, the company is planning to add support for NVIDIA’s Nemotron model and DeepSeek V4 Flash to PCs that are running on Arm-based RTX Spark devices, which are slated to start shipping this fall from Microsoft and various PC makers.
Execs also demonstrated how the new Copilot app potentially could make AI more useful to Windows users. Rather than adding gratuitous Copilot buttons to various parts of Windows, as was Microsoft’s plan not long ago, the new Copilot app will be able to access files, email, and other local information — with users’ permission — and use that information to tailor prompt results. Via the coming Microsoft “Autopilot” long-running agent capability, Copilot will be able to go a step further and take action on Windows via tasks like organizing files, troubleshooting issues and more.
Autopilot, not to be confused with Windows Autopilot, is the renamed OpenClaw-based Microsoft Scout personal agent. It is in private preview. Unsurprisingly, officials did not mention that Autopilot will require its own separate license, as it will be provisioned like a human, as Microsoft has said in the past.
Microsoft also announced on Oct. 7 several new Windows PCs built to support local AI processing that users can now pre-order new Windows PCs powered by the Arm-based NVIDIA RTX Spark from ASUS, Dell, HP, Lenovo, MSI and Microsoft Surface. The Surface Laptop Ultra starts at $2,600 and Surface RTX Spark Dev Box starts at $6,000 USD.
