Microsoft proposes 'hybrid intelligence' for Windows 11, spanning local and cloud processing

At a San Francisco event on October 7, Microsoft introduced its "hybrid intelligence" approach, a mixed architecture that combines cloud AI and local computing to handle tasks on Windows 11.

According to Microsoft, Windows AI will no longer rely solely on cloud or on-device processing. Instead, it will coordinate between local and cloud based on the task type, device performance, and security requirements: privacy-sensitive or latency-sensitive tasks can rely more on local processing, while complex models and heavy compute workloads can run in the cloud.

Windows 11 is expected to be a key entry point for this model. After a user defines a goal, Copilot and AI agents are responsible for understanding context, invoking tools, and advancing the task, while the user retains control over critical steps, permissions, and final outcomes. Microsoft CEO Satya Nadella described trust, intelligence, and "AI that serves human ambition" as three pillars, and called Agent 365, Microsoft IQ, and Copilot key levers.

Microsoft also highlighted the importance of local-cloud coordination, with on-device processing suited for low-latency and privacy-sensitive tasks and the cloud providing stronger models and computing power. Together with controlled execution environments such as Execution Containers, AI agents could perform more complex actions when granted the necessary permissions. During the event, Microsoft senior technical program manager Kayla Cinnamon demonstrated GitHub Copilot automatically selecting a local model rather than defaulting to cloud models.