Microsoft has drawn a sharp line in the sand for Windows 11 laptops. Every recent machine can run the Copilot AI assistant, but only a select few qualify as Copilot Plus PCs—and the difference comes down to one number: 40 trillion operations per second. That’s the threshold for the neural processing unit, or NPU, a specialized chip now required to unlock the full promise of AI on Windows.

For months, the term “AI PC” has been tossed around loosely, but Copilot Plus gives it a precise definition. It’s not just about having an AI chatbot in the taskbar; it’s about running advanced machine learning workloads locally, without melting your battery or clogging the cloud. Microsoft’s vision, unveiled in May 2024, ties this new category to strict minimums: a processor with an NPU capable of at least 40 TOPS, 16GB of RAM, and 256GB of storage. If your laptop doesn’t meet all three, it’s simply not a Copilot Plus PC.

What Is a Neural Processing Unit and Why TOPS Matter

TOPS, or trillions of operations per second, is the de facto speedometer for AI silicon. Unlike a GPU, which excels at parallel floating-point math for graphics, an NPU is built from the ground up for the matrix multiplications and convolutions that neural networks devour. A high TOPS count means the NPU can juggle larger models or generate responses faster—think real-time language translation, background blur that actually looks natural, or creative image generation without a network hiccup.

Before Copilot Plus, most Windows NPUs were in the 10–15 TOPS range. Intel’s Meteor Lake, for instance, integrates an NPU rated at about 11 TOPS. That’s enough for subtle Studio Effects in video calls, but not for the kind of deeply integrated AI Microsoft envisions. The 40 TOPS floor isn’t arbitrary; it’s the level Microsoft determined necessary to run multiple AI models concurrently while leaving headroom for the operating system and applications. With 40+ TOPS, an NPU can handle heavy inference—like powering the controversial Recall feature, which continuously screenshots your activity to make it searchable—without dragging down the CPU or GPU.

The Hardware Trinity: NPU, RAM, and Storage

Copilot Plus PCs aren’t just about big numbers on a spec sheet. The 16GB RAM requirement ensures that large language models, which can easily gobble up 4–6GB of memory, have enough headroom. Run an AI model alongside a few browser tabs and Slack, and 8GB machines quickly hit a wall. Microsoft’s floor acknowledges that AI features are memory-hungry; 16GB is the new baseline for a premium Windows experience.

The 256GB storage requirement might raise eyebrows. After all, cloud-based Copilot doesn’t care about your SSD. But local AI models need local storage, and some features—like Windows Studio Effects packages or offline language packs—can be hefty. Moreover, Microsoft’s Recall feature stores a rolling database of screenshots, which could grow to tens of gigabytes over time. Fast NVMe storage is also critical because AI workloads often stream large tensors from disk if they don’t fit in RAM.

At launch, the only silicon meeting the 40+ TOPS bar was Qualcomm’s Snapdragon X Elite and Snapdragon X Plus platforms, which leverage an Arm-based CPU paired with a custom Hexagon NPU. These chips deliver up to 45 TOPS, giving Snapdragon-powered laptops an early lead. Intel and AMD are racing to catch up: Intel’s Lunar Lake and AMD’s Strix Point processors, both announced later, will also incorporate NPUs exceeding 40 TOPS. So while the first wave of Copilot Plus PCs is exclusively Arm, the category will soon embrace x86.

Copilot vs. Copilot Plus: Feature Gaps and User Experience

Every Windows 11 laptop with a recent update can summon the Copilot sidebar. It’s a powerful web-connected assistant that can rewrite text, summarize documents, and answer questions by tapping into Bing and Microsoft 365. But Copilot Plus PCs unlock a set of on-device AI experiences that never leave your hardware:

  • Recall: A photographic memory for your PC. Recall captures snapshots of your screen every few seconds, then uses on-device AI to make that content searchable. Want to find that slide deck you saw three weeks ago? Type a vague description and Recall resurfaces it. This feature is exclusive to Copilot Plus because it demands constant, efficient AI processing that would annihilate battery life on a weaker NPU.
  • Cocreator in Paint: The updated Paint app uses a local diffusion model to turn your doodles into polished artwork in real time. On Copilot Plus PCs, this happens instantly; on older devices, it would rely on cloud inferencing, introducing latency and privacy concerns.
  • Live Captions with Translation: Copilot Plus PCs can generate captions for any audio flowing through the system—videos, meetings, podcasts—and translate them into English from dozens of languages, all offline.
  • Windows Studio Effects Enhanced: Background blur, eye contact correction, and automatic framing are sharper and more responsive, with new effects like portrait light and creative filters that run continuously during video calls.
  • Enhanced Security with the Microsoft Pluton processor: Copilot Plus PCs require a Pluton security chip, which hardens the device against firmware attacks and protects encryption keys, including those used for AI data.

Beyond features, the NPU also offloads AI tasks that would otherwise bog down the CPU and GPU. This means longer battery life—Snapdragon X Elite laptops routinely hit 15–20 hours on a charge, partly because the NPU handles AI whisper-quietly in the background. And because processing happens locally, sensitive data never traverses the internet, a boon for privacy-conscious users and enterprises.

On a standard Copilot laptop, many of these features either don’t exist or run in a diminished form. Copilot remains essentially a cloud service; you’re not getting on-device Recall or Cocreator. The assistant is still valuable, but it’s tethered to Microsoft’s servers and your internet connection.

The Snapdragon X Factor: Architecture and App Compatibility

The first Copilot Plus PCs are all powered by Snapdragon X chips, which means they run Windows on Arm. That architectural shift brings its own baggage: while Microsoft has heavily invested in an emulation layer called Prism, some legacy x64 apps may run slower or not at all. Creative pros reliant on certain Adobe plugins or niche engineering software should verify compatibility, though the situation is improving rapidly. Major applications like Microsoft Office, Edge, Chrome (via a native Arm version), Photoshop, and DaVinci Resolve all have Arm-native builds. The performance of the Snapdragon X Elite in general computing is competitive with Apple’s M3, and battery life often surpasses Intel-based ultrabooks.

The Arm-to-x86 transition isn’t new—Apple’s M-series proved it can be seamless—but Windows users may experience friction. The good news: for most everyday tasks, including AI workloads, the experience is snappy and transparent. And as Intel and AMD ship their own 40 TOPS NPUs, buyers will have a choice between architectures without sacrificing Copilot Plus benefits.

Who Should Buy a Copilot Plus PC?

If you’re content with the current Copilot sidebar and don’t need AI to be woven into the operating system’s fabric, a regular Windows 11 laptop is perfectly serviceable. Copilot will continue to improve on the cloud side, and many exciting generative AI features—like Microsoft 365 Copilot in Word and Excel—run server-side anyway.

But if you want to future-proof your machine for an era where AI is as fundamental as a keyboard and touchscreen, a Copilot Plus PC is the smarter bet. Microsoft has signaled that forthcoming Windows updates will increasingly lean on local AI. Features like Recall may become central to how we interact with our digital histories. Developers are already leveraging the Windows Copilot Runtime, a set of APIs that tap directly into the NPU, to build AI-accelerated apps. Over the next two years, the software ecosystem will bifurcate: apps that require an NPU and apps that don’t.

For business users, Copilot Plus PCs offer a privacy advantage. Data never leaves the device for many AI tasks, easing compliance with regulations like GDPR and HIPAA. IT departments can confidently deploy Recall knowing that sensitive company information isn’t being uploaded to Microsoft’s cloud (Recall’s snapshots are encrypted and stored locally).

The Road Ahead: Intel, AMD, and the Mainstreaming of AI PCs

Intel’s Lunar Lake is expected to hit the market with over 40 TOPS, bringing Copilot Plus to the x86 ecosystem. AMD’s Strix Point similarly boasts an NPU that exceeds the threshold. By the 2024 holiday season, consumers will likely have dozens of Copilot Plus PC models to choose from, spanning various price points and form factors.

Competition will drive down costs. Today, the entry price for a Copilot Plus PC hovers around $999 with the Snapdragon X Plus, but that will fall as the technology trickles into mainstream laptops. Crucially, the NPU and RAM minimums are tied to the Copilot Plus branding, not to Windows 11 itself; you’ll still be able to run Windows 11 on cheaper, less powerful hardware. But the most compelling AI features will be locked behind that hardware gate.

Conclusion

The distinction between Copilot and Copilot Plus PC marks a turning point for Windows laptops. It’s the difference between a simple AI assistant tacked onto the OS and a deeply integrated, always-on co-pilot that anticipates your needs. The 40 TOPS NPU is the engine of that transformation, enabling a new class of experiences that are fast, private, and efficient. Whether you jump on the Arm-powered Snapdragon X bandwagon now or wait for Intel and AMD to join the party, one thing is clear: the era of the AI PC just got real, and the requirements are more than marketing fluff. They’re the new baseline for what a premium Windows laptop should be.