A dedicated neural processing unit (NPU) handles AI inference tasks, like live captioning or background indexing, at a fraction of the power draw the CPU or GPU would use for the same workload, because it's purpose-built silicon rather than general-purpose compute pressed into service. That efficiency gain is real, but it doesn't mean AI features are automatically free: a feature that runs continuously in the background, even efficiently, still adds some drain across a full day compared to having it switched off entirely.
How NPUs Change the Power Equation
Before dedicated NPUs, a laptop running any AI-adjacent task, like real-time noise cancellation on a call, pushed that work onto the CPU or integrated GPU, both of which draw meaningfully more power per operation for that specific kind of math. An NPU is designed only for that narrow category of workload and does it at much lower wattage, which is the entire reason manufacturers added the chip rather than just using the CPU.
Which AI Features Actually Run Continuously
Background indexing features like Windows Recall, if enabled, periodically process activity in the background rather than only when you actively request it, meaning they draw some power even when you're not consciously using them. Features like Cocreator or on-demand image generation, by contrast, only draw power when you actively invoke them, closer to opening any other app briefly.
Checking What's Actually Draining Your Battery
Go to Settings > System > Power & battery > Battery usage to see a per-app and per-feature breakdown of what's consumed battery over the last day or week. This is more useful than guessing, since it's common to assume an AI feature is the culprit when a browser tab or background sync app is actually the bigger drain.
When Disabling AI Features Genuinely Extends Battery Life
If Battery usage shows a background AI indexing feature consistently near the top of the list, disabling it (Settings > Privacy & security > Recall & snapshots, for example) will measurably extend runtime on battery, particularly useful before a long day away from a charger. For occasional-use features that only run when invoked, disabling them saves little since they weren't drawing power continuously anyway.
Battery Health Still Matters More Than Any Single Feature
A laptop with a healthy battery handles the modest added draw from AI features without much visible impact on all-day usability, while a laptop with a battery already at 60-70% of original capacity will feel every added drain far more acutely. If your laptop's runtime has dropped sharply regardless of which features you enable or disable, the battery itself, not the software, is likely the real cause.
A Technical Check: Battery Report and Efficiency Mode
Run powercfg /batteryreport in Command Prompt to see your battery's current capacity against its original design capacity, the clearest sign of whether wear, not software, explains a runtime drop. Task Manager's Efficiency mode (right-click any process) can also force a background app to use fewer resources without closing it outright, useful for AI-adjacent background apps that don't have their own power setting. If the battery report shows significant degradation, a battery replacement addresses the actual cause far more effectively than disabling features one by one.
