What does an NPU do and why do so many laptops have them? (www.engadget.com)

Recent reports indicate that many new laptops are equipped with Neural Processing Units (NPUs), specialized chips designed to handle AI tasks efficiently. While these chips consume significantly less power—around 2W compared to over 100W for traditional GPUs—they remain underutilized. The push for NPUs is largely driven by Microsoft's new certification requirements, which mandate that new laptops achieve at least 40 TOPS (Tera Operations Per Second) for NPU performance, positioning them for enhanced Windows AI features. This has led to a surge in NPU availability across mid-range and premium laptops, even though applicable software is still scarce. NPUs are intended primarily for AI inference, allowing for power-efficient execution of tasks like background blur during video calls without draining battery life. However, they are not designed for training AI models and fall short for demanding applications like large language models. As manufacturers like Qualcomm, Intel, and AMD continuously increase NPU capabilities—some models now exceeding 85 TOPS—the technology is set to become a standard feature in laptops, potentially making NPUs indispensable for future AI applications. Users can check NPU utilization through Windows Task Manager, though many casual users may find their NPUs largely idle due to limited software support.