What Is a Copilot+ PC? Meaning and Requirements Explained

A Copilot+ PC is a Windows 11 laptop (or 2-in-1) that meets Microsoft's hardware certification for on-device AI: a neural processing unit (NPU) rated at 40 or more trillion operations per second (TOPS), at least 16GB of RAM, and at least 256GB of storage. Meeting that bar unlocks a set of Windows features — Recall, Click to Do, improved semantic search, on-device image generation, and more — that simply won't run, or won't run well, on a regular Windows 11 machine. It's a certification and marketing badge, similar in spirit to "Intel Inside" or Energy Star, not a single product or a single chipmaker's technology.

What is a Copilot+ PC, exactly?

Microsoft introduced the Copilot+ PC category in May 2024 as a new tier of Windows 11 hardware built specifically to run AI models locally, on the device, instead of sending every request to the cloud. The company's own announcement described Copilot+ PCs as "the fastest, most intelligent Windows PCs ever built," powered by silicon capable of 40-plus TOPS, with all-day battery life and access to large AI models that run directly on the machine. You can read Microsoft's original pitch in full on the Official Microsoft Blog.

The badge is a certification, not a product. A "Copilot+ PC" can be a Surface, a Dell XPS, a Lenovo ThinkPad, an HP, an Asus, an Acer, or a Samsung Galaxy Book — any laptop from any of Microsoft's hardware partners, as long as it clears the NPU, RAM, and storage bar Microsoft sets. The same way "Energy Star" doesn't name a single appliance, "Copilot+ PC" doesn't name a single laptop.

Copilot+ PC requirements: NPU, RAM, and storage

Microsoft lists the exact hardware floor on its own Windows 11 specifications page. To carry the Copilot+ PC badge, a device needs:

• An NPU capable of 40+ TOPS
• 16GB of DDR5 or LPDDR5 RAM
• 256GB of SSD or UFS storage

All three thresholds have to be met together. A laptop with a 45 TOPS NPU but only 8GB of RAM doesn't qualify, and neither does a 16GB, 512GB-storage laptop whose NPU tops out at 20 TOPS. Microsoft is also explicit that the TOPS figure has to come from the NPU specifically — not the CPU, not the GPU, and not a combined "platform TOPS" number that chipmakers sometimes advertise, which typically sums NPU, GPU, and CPU AI throughput together. A gaming laptop with a powerful discrete GPU can often exceed 40 TOPS of AI compute overall, but if that horsepower is coming from the graphics card rather than a dedicated NPU, the machine still isn't a Copilot+ PC, because the GPU draws far more power to get there and can't run Copilot+ features continuously in the background the way an efficient NPU can.

RAM matters for the same reason: on-device language models and image generators keep multi-gigabyte models resident in memory, and Recall's continually growing semantic index needs headroom alongside whatever else is open. Storage capacity matters because Recall's snapshot index and the on-device models Windows downloads both consume real disk space over time. If you're shopping for a Copilot+ laptop and weighing RAM configurations, it's also worth knowing where memory technology is headed next — see our explainer on DDR6 RAM for how the next memory generation compares to the DDR5/LPDDR5 that every current Copilot+ PC uses.

What is an NPU, and why does TOPS matter?

A neural processing unit (NPU) is a dedicated chip, built into the processor package, whose only job is running the matrix-multiplication math that neural networks are made of. Unlike a CPU, which is a generalist built to execute almost any instruction one (or a handful of) at a time, an NPU is a specialist: it's wired to do enormous numbers of small, repetitive, low-precision calculations — the kind that happen millions of times over when a model processes an image, a sentence, or a frame of audio — in parallel, and to do it using a fraction of the electricity a CPU or GPU would burn on the same task.

TOPS stands for trillion operations per second, and it's the standard yardstick for an NPU's raw throughput, usually measured using 8-bit integer (INT8) math, the precision most on-device AI models are optimized to run at. More TOPS generally means a chip can run larger models, run them faster, or run several AI features at once without stuttering. Microsoft's 40 TOPS line isn't an arbitrary round number — it's the level Microsoft and its silicon partners determined was necessary to run features like Recall's continuous indexing and live, offline Windows Studio Effects without noticeably draining the battery or heating up a thin laptop chassis. The original announcement describes the NPU as the component that lets these experiences run "locally, at the edge," rather than depending on a cloud connection — see the architecture explanation on the Microsoft Blog.

NPU vs. GPU: why AI work moved to a third processor

If your laptop already has a GPU, and GPUs are good at the parallel math AI needs, why does Windows need a whole separate NPU? Three reasons:

• Power efficiency. A discrete GPU can deliver far more raw AI throughput than a current NPU, but it does it by drawing tens to well over a hundred watts. An NPU delivers its 40-plus TOPS at a few watts, which is the difference between a feature that quietly runs all day in the background and one that would drain a laptop battery in under an hour.
• Thermals. A GPU running flat-out generates heat that thin-and-light laptops, the kind Copilot+ PCs are mostly built as, aren't designed to dissipate continuously. An NPU is built for sustained, always-on background work rather than short bursts.
• Availability. Not every Windows laptop has a discrete GPU — most thin, affordable ones rely on integrated graphics only. Building AI acceleration into the chip package itself, the way CPUs and now NPUs are, guarantees every Copilot+ PC has the same baseline capability regardless of whether it has dedicated graphics.

In short: GPUs are still the right tool for gaming and heavy creative rendering, CPUs are still the right tool for general-purpose code, and the NPU is the tool built specifically to run AI inference efficiently enough to leave running in the background, all day, on battery.

Which chips and laptops qualify as Copilot+ PCs

As of late 2026, three silicon families meet the Copilot+ PC bar, from all three Windows chip vendors:

Qualcomm's original Snapdragon X Elite and X Plus were the launch chips for the category in 2024, at 45 TOPS. The newer Snapdragon X2 Elite and X2 Elite Extreme, announced by Qualcomm in September 2025, nearly double that figure: Qualcomm's own release describes the Hexagon NPU in these chips as delivering 80 TOPS of AI processing, which the company calls "the world's fastest NPU for laptops," per Qualcomm's press release. For a closer look at how Qualcomm's chip stacks up against Intel's in day-to-day use, see our Snapdragon X2 Elite vs. Intel Core Ultra 300 comparison.

Intel's Lunar Lake (Core Ultra 200V series) was Intel's first Copilot+-qualifying chip, launched in late 2024 at 48 TOPS. Its successor, Panther Lake — branded Core Ultra Series 3, sometimes written Core Ultra 300 — launched at CES 2026 on Intel's 18A manufacturing process. Intel's own newsroom describes Panther Lake as delivering up to 180 TOPS of combined platform AI performance across the CPU, GPU, and NPU, with the dedicated NPU comfortably clearing Microsoft's 40 TOPS floor on its own. Full specs and release timing are covered in our Panther Lake release date and specs explainer.

AMD's Ryzen AI 300 series (XDNA 2 NPU, up to 50 TOPS) has powered Copilot+ laptops since mid-2024, and the Ryzen AI 400 series extended the same 50 TOPS NPU to AMD's newest mobile chips and, for the first time, to desktop processors — AMD calls the Ryzen AI 400 series "the world's first" desktop chips built to support Copilot+ experiences, according to AMD's newsroom.

Because all three vendors now clear 40 TOPS comfortably, the gating factor for most current laptops is RAM and storage configuration, not NPU performance. A cheap configuration of an otherwise-qualifying chip, sold with 8GB of RAM, still won't carry the Copilot+ PC badge.

What features does a Copilot+ PC actually unlock?

Per Microsoft's own Windows 11 specifications page, the following experiences are gated behind full Copilot+ PC hardware compliance and won't appear, or won't function, on a standard Windows 11 laptop:

• Recall — a searchable, locally-stored timeline of what you've seen and done on the PC, built from periodic encrypted snapshots, with Windows Hello Enhanced Sign-in Security required to unlock it
• Click to Do — a system-wide assistant that suggests contextual actions (summarize, rewrite, look up) based on whatever is on screen
• Improved Windows search — semantic, natural-language search across local files and settings
• Paint Object select / Generative Fill and Photos Super resolution / Restyle Image — on-device AI image editing
• Snipping Tool's "Perfect screenshot" and Live Captions real-time translation — processed locally rather than in the cloud
• Cocreator in Paint — near-real-time AI image generation from sketches and text prompts

A few features go further and are currently exclusive specifically to Snapdragon-powered Copilot+ PCs, including Automatic Super Resolution for upscaling older games and some Photos relighting tools, reflecting how Arm-based Snapdragon silicon was first out of the gate with the category. Intel and AMD Copilot+ PCs get the rest of the core feature set.

Copilot+ PC vs. "AI PC" vs. a regular Windows 11 laptop

These three terms get used almost interchangeably in marketing, but they mean different things:

• A regular Windows 11 laptop has no NPU requirement at all. It can still run cloud-based Copilot and other AI apps over the internet, but nothing runs on-device, and none of the Copilot+ exclusive features are available.
• An "AI PC" is a looser, industry-wide marketing term (used by Intel, AMD, and PC makers generally) for any laptop with some NPU on board, regardless of how fast it is. A machine with a 10 or 20 TOPS NPU is routinely marketed as an "AI PC" but falls well short of Microsoft's 40 TOPS Copilot+ threshold.
• A Copilot+ PC is Microsoft's specific, certified tier within "AI PC": 40+ TOPS NPU, 16GB RAM, 256GB storage, verified by Microsoft, which is what unlocks Recall, Click to Do, and the rest of the exclusive feature list.

Put simply, every Copilot+ PC is an AI PC, but not every AI PC is a Copilot+ PC. If a listing just says "AI PC" without a specific TOPS figure or the Copilot+ badge, assume it doesn't meet the certification and won't run Recall or Click to Do.

How much does a Copilot+ PC cost?

Microsoft's May 2024 launch lineup — the first wave of Snapdragon-powered Copilot+ PCs, including Surface devices and OEM machines from Acer, Asus, Dell, HP, Lenovo, and Samsung — started at $999, as stated in Microsoft's own announcement on the Official Microsoft Blog. Pricing has broadened considerably since: Intel and AMD Copilot+ configurations have appeared at a range of price points as the qualifying chips spread across more of each vendor's lineup, and premium configurations — like Microsoft's own high-end Surface hardware — sit well above that $999 starting point.

Should you buy a Copilot+ PC, and what's next

If you're buying a new Windows laptop today, a Copilot+ PC is rarely a downgrade: the NPU, RAM, and storage floor Microsoft requires also tends to produce a genuinely well-specced, efficient machine even if you never touch Recall or Click to Do, and battery life on Copilot+ class silicon is generally strong across all three chip vendors. Whether the exclusive features are worth paying a premium for is more personal — Recall in particular remains a feature some buyers actively want and others prefer to leave switched off, since it indexes your own on-screen activity even though that index stays local to the device.

Looking ahead, expect the Copilot+ bar itself to keep moving. Every qualifying chip family already clears 40 TOPS by a wide margin — Qualcomm's newest silicon is roughly double the minimum — so it's plausible Microsoft raises the NPU floor again as the installed base of higher-TOPS chips grows, the same way the original 2024 minimum specs got incrementally expanded to cover more Intel and AMD silicon over time. For now, the fastest-growing part of the story is less about any single chip and more about connectivity and memory keeping pace with these faster NPUs.

Written byMara Lindqvist

Mara Lindqvist edits the hardware desk. She covers graphics cards, processors, memory and storage, the foundries and chip designers behind them, and what the numbers on a spec sheet mean for people choosing a PC. Specifications in her stories come from manufacturer spec pages and datasheets.

More from Mara LindqvistAll Hardware coverage