3 Minutes
Microsoft says a smarter Windows is coming, powered by tiny AI chips called NPUs. These neural processing units promise faster on-device AI, longer battery life and new agent-driven interactions that could change how we use our PCs.
NPUs: small chips, outsized impact
Neural Processing Units (NPUs) are specialized chips designed to run machine learning tasks far more efficiently than general-purpose CPUs or GPUs. Microsoft highlights NPUs that can deliver roughly 40 trillion operations per second (40 TOPS), enabling sophisticated AI features to run locally without a constant cloud connection.
That efficiency matters. By offloading AI work to an NPU, PCs can offer instant, private experiences while sipping battery power — and the hardware required fits into machines that cost only a few hundred dollars, not server racks.
Why Microsoft calls them Copilot+ PCs
Microsoft has been promoting the Copilot+ PC idea for over a year. In their vision, these devices combine NPUs with cloud services to balance local responsiveness and large-model capabilities. Small models run locally on the NPU for quick tasks like recall, transcription or camera processing, and heavier workloads can fall back to cloud models when needed.
One origin story for this approach is Microsoft’s Surface Hub 2 Smart Camera, which ran AI tasks on-device efficiently. Lessons from that project were shared with chip partners such as AMD, Qualcomm and Intel, who then started building PC-focused AI silicon.
Agents, fewer clicks, more context
Steven Bathiche, who leads Microsoft’s Applied Sciences team, calls these agentic experiences the new unit of interaction. In plain terms: Windows is moving toward workflows that do things for you rather than only letting you click through menus. Microsoft already teases a dedicated agent in Settings, and says more complex automations are coming.
Imagine asking your PC to summarize a week of emails, schedule time for follow-ups and surface the most important threads — all with minimal clicking. That is the kind of experience NPUs help make practical, because local processing provides speed and privacy while cloud services supply heavier reasoning.
What this means for users and developers
- Users get faster, more private AI features on mainstream hardware.
- Developers can target agent-based workflows that combine local and cloud models.
- Hardware makers are racing to add NPUs to more laptop and desktop designs.
Microsoft frames NPUs as critical infrastructure for the next generation of AI in Windows. By blending local neural processing with cloud intelligence, Copilot+ PCs aim to deliver smarter computing without pushing everyone into expensive, specialized devices.
Source: neowin
Leave a Comment