Microsoft dropped Windows 11 Insider Preview build 28020.1921 on May 1, 2026, handing its Experimental (26H1) channel a long-awaited feature: live NPU activity tracking directly in Task Manager. A second build, 29580.1000, also shipped for Future Platforms, signaling aggressive testing for unannounced hardware. But it’s the Task Manager tweak that’s turning heads — and for good reason.

The new column, tucked inside the Processes tab, exposes Neural Processing Unit utilization for each running application. No more guesswork. No more third-party tools. With a single glance, developers and enthusiasts can see if an app is actually tapping the dedicated AI silicon inside their PC — or just pretending to.

What the New Builds Actually Include

Build 28020.1921 is tagged as an Experimental (26H1) release. Microsoft uses the 26H1 moniker to denote feature updates expected in the first half of 2026. Insiders in the Experimental channel get early, often unstable, previews of upcoming functionality. This build’s headliner is the Task Manager NPU column, but the full changelog remains under wraps. Microsoft typically discloses more details in its blog post, which was not yet live at the time of writing.

Simultaneously, build 29580.1000 landed in the Experimental (Future Platforms) ring. That’s a decidedly more mysterious branch. The 29000-series build numbers point toward next-generation hardware support — likely Copilot+ PCs with Qualcomm’s Snapdragon X Gen 2 or Intel’s Core Ultra 300 series. Early testers in this channel are probing driver compatibility, firmware integration, and OS-level AI pipelines that will define Windows 11’s evolution through 2027 and beyond.

Both builds are limited to Windows Insiders who opted into the Experimental tier, meaning general availability is still months away. Users can check their enrollment via Settings > Windows Update > Windows Insider Program.

Inside the Task Manager NPU Column

For years, Task Manager has been the go-to performance dashboard — CPU, GPU, memory, disk, network. But as discrete AI engines become standard on everything from $300 laptops to $3,000 workstations, Microsoft faced a blunt reality: the tool was blind to one of the most critical subsystems in modern PCs.

That blindness ends today.

The new NPU column functions identically to the CPU or GPU columns. It displays a percentage representing current NPU load per process. If an app isn’t using the NPU, the column shows “0%” or is blank. If a video conferencing tool is running real-time background blur via Windows Studio Effects, you’ll see the corresponding spike.

Behind the scenes, the Operating System hooks into the Windows NPU Management API, which marshals work across heterogeneous compute units. Developers targeting the Windows Copilot Runtime can register their models with the NPU scheduler. Task Manager then queries usage counters exposed by the driver. The result is real-time, low-overhead monitoring.

Key behaviors to expect:
- The column appears by default only on systems with a detected NPU. Microsoft has baked hardware detection into the shell experience.
- You can add or remove the column via right-click on the column header, just like any other metric.
- The percentage reflects total NPU utilization, not core-level breakdown. NPUs often contain multiple compute tiles, but the OS abstracts that complexity.
- Background processes like Windows’ own AI indexing or captioning services may register activity, giving users new insight into system resource consumption.

Early reports from Insiders suggest that the column works reliably on Snapdragon X Elite and Intel Meteor Lake systems — hardware already prevalent in the 2024-2025 Copilot+ wave. Future silicon with enhanced Hexagon, Movidius, or Intel GNA blocks should plug in without additional driver updates.

Why an NPU Column Matters More Than You Think

This isn’t just a cosmetic tweak. It’s a foundational shift in how users interact with AI workloads.

1. Transparency for power users
When a laptop’s battery tanks during a Teams call, users can now pinpoint the culprit. Is it the GPU handling video decoding? The CPU wrestling with network interrupts? Or the NPU chewing through background segmentation? Task Manager finally closes that diagnostic loop.

2. Developer accountability
App developers have hyped “AI acceleration” in marketing slides for years. Now users can verify those claims directly. If a photo editor touts NPU-powered upscaling but the column stays flat, something’s amiss. This visibility pressures ISVs to optimize their AI pipelines properly.

3. AI PC validation
Microsoft and its OEM partners have sold the idea of the “AI PC” aggressively. But until now, measuring AI compute was cumbersome. The NPU column provides an objective, system-wide metric. Reviewers can include Task Manager screenshots in benchmarks. IT admins can audit fleet usage. The narrative shifts from abstract to empirical.

4. Educational value
Most users have never heard of an NPU. Seeing a new column next to CPU and GPU normalizes the concept. It signals that AI processing is as fundamental as graphics rendering. That’s a small but important step toward mainstream AI literacy.

The Bigger Picture: Windows as an AI Operating System

Task Manager’s NPU column is the tip of a very large iceberg. Microsoft has been retooling Windows around artificial intelligence since at least 2023. The roadmap includes:

  • Windows Copilot Runtime: A set of APIs and services that let apps leverage on-device models for text, image, and audio tasks.
  • Recall (AI Timeline): Semantic search across everything you’ve ever done on your PC, powered by local models running on the NPU.
  • Studio Effects: Real-time camera and microphone processing, shipped as part of the OS, using the NPU to offload work from CPU/GPU.
  • AI Explorer: A contextual assistant that understands screen content and suggests actions, again leaning heavily on NPU inference.

All these features demand a hardware-aware monitoring solution. Task Manager’s update closes the feedback loop between the OS’s AI ambitions and the user’s understanding.

Furthermore, Microsoft’s silicon strategy is diversifying. The Future Platforms build (29580.1000) nods to custom silicon — perhaps the rumored “Azure Maia” chip adapted for client devices, or a refreshed HoloLens processor. By decoupling build numbers from specific hardware, Microsoft can validate multiple chipset families in parallel. Insider telemetry will reveal which NPU architectures deliver the best performance-per-watt for common Windows AI tasks.

What Insiders Are Saying

Early feedback from the Windows Insider community has been predominantly positive. On the WindowsForum (a prominent hub for Windows enthusiasts), one tester noted: “Finally, I can see if my $1500 ‘AI PC’ is actually doing AI stuff. Spoiler: so far, only Teams and Photoshop are hitting the NPU.”

Others have raised valid concerns:

  • Granularity: The column shows overall NPU usage, not which part of the chip (e.g., scalar vs. vector units) is active. For deep tuning, developers still need vendor-specific profiling tools.
  • Driver maturity: On some older Intel Meteor Lake systems, the NPU column sporadically disappears after wake-from-sleep. Driver updates will likely address this.
  • Cross-process aggregation: If multiple background services use the NPU simultaneously, the per-process breakdown can be less intuitive than a unified “NPU total” graph. Microsoft may add a Performance tab graph in future builds.

Nonetheless, the consensus is clear: this is a necessary, overdue addition.

How to Get the Feature

If you’re itching to test the NPU column, here’s the step-by-step:

  1. Join the Windows Insider Program: Open Settings, navigate to Windows Update > Windows Insider Program, and sign in with a Microsoft account.
  2. Choose the right channel: Select “Experimental” to receive 28020.1921. Note that Experimental builds can be buggy — don’t install on a production machine.
  3. Have an NPU-equipped PC: The column only appears on devices with a dedicated NPU. Most Copilot+ laptops released after 2024 qualify, including the Surface Pro 10 and Lenovo Yoga Slim 7x.
  4. Check for updates: After enrolling, go to Windows Update and click “Check for updates.” Build 28020.1921 should download automatically.

If you don’t see the column, verify your NPU driver via Device Manager. The NPU typically appears under “Neural Processors” or “System Devices,” depending on the vendor.

The Road Ahead

Microsoft hasn’t publicly committed to shipping the NPU column in the final Windows 11 26H1 update, but Insider features with this level of polish rarely get scrapped. The 26H1 update is expected to roll out in October 2026, aligning with new hardware launches from Intel and AMD.

More intriguing is the Future Platforms build. If Microsoft is already testing non-mainstream silicon, we could see a Windows variant tailored for ARM64 workstations, AI inference accelerators, or even automotive applications. The unified Windows Core OS strategy makes such flexibility plausible.

What’s missing, for now, is a counterpart in the Performance tab. A real-time NPU graph would help visualize spikes over time. Similarly, resource monitor integration is absent. Power users will likely clamor for these additions, and Microsoft’s Insider channels provide exactly that feedback loop.

Bottom Line

Windows 11 Insider build 28020.1921 transforms Task Manager from a passive observer into an AI-aware command center. By surfacing NPU usage alongside CPU, memory, disk, and GPU, Microsoft acknowledges that neural processing is now a first-class citizen in the Windows ecosystem. The companion build 29580.1000 hints at a future where even more exotic hardware will be managed through the same lens.

For enthusiasts and developers, the message is unmistakable: the AI PC is real, and it’s time to start monitoring it like one.