The Windows PC landscape took a decisive turn at CES 2026 as Intel, AMD, and Qualcomm each shipped—or at least formally dated—next-generation processors with dedicated neural processing units (NPUs) powerful enough to run Microsoft’s Copilot+ experiences locally. New laptops and radical form factors, including a full Windows PC stuffed inside a keyboard, moved on-device AI from a talking point to a purchasing requirement.
What actually changed at CES 2026
Three chip giants fielded silicon that finally crosses the performance threshold Microsoft set for Copilot+ certification. Intel’s Core Ultra Series 3 (code-named Panther Lake) became the first consumer platform built on the company’s U.S.-based 18A manufacturing process, hitting up to 50 trillion operations per second (TOPS) on its integrated NPU in top SKUs. Pre-orders opened January 6, with broad laptop availability confirmed for January 27, 2026, according to Intel’s press materials.
AMD answered with the Ryzen AI 400 family, a Zen 5 and XDNA 2 NPU design that pushes up to 60 TOPS in its fastest configurations. The chips will populate thin-and-light laptops through to workstation-class machines, with all major OEMs showing immediate support.
Qualcomm’s Snapdragon X2 Plus stole attention by bringing an 80 TOPS Hexagon NPU—the same figure found on the pricier X2 Elite—to a mid-tier Arm processor. The move effectively democratizes high-end AI throughput for more affordable Windows on Arm devices, something that could reshape the budget and mainstream segments if software compatibility keeps pace.
Simultaneously, PC makers unveiled hardware that makes these NPUs tangible. Acer’s Swift 16 AI features what the company calls the world’s largest haptic touchpad and runs on an Intel Series 3 chip. ASUS doubled down on its dual-screen Zenbook Duo, now with an on-device NPU for real-time translation and content generation. Dell revived the XPS 14 and 16 with aluminum CNC bodies, better thermals, and a new 6K UltraSharp monitor for creators. Lenovo and Samsung refreshed their ThinkPad and Galaxy Book lines with Copilot+ badges and AI-forward hinges. But the show-stealer was HP’s EliteBoard G1a, a keyboard PC powered by an AMD Ryzen AI 300-series processor with an NPU rated above 50 TOPS. HP says it will be available to enterprise customers in March 2026.
Every major reveal shared a common thread: a dedicated NPU that meets or exceeds the 40 TOPS baseline Microsoft defines for Copilot+ experiences, making local AI acceleration a standard feature rather than an optional add-on.
What it means for you, whether you’re buying for home, work, or coding
For home users and creators
AI-assisted photo and video editing, background blur in video calls, and real-time language translation all become faster and more private because the processing stays on your device. Arm-based laptops like those with Snapdragon X2 Plus promise multi-day battery life while still handling Copilot tasks without a cloud round-trip. The catch: not every app will see an instant speed boost. Performance relies on software being written or updated to leverage the NPU, so you’ll want to check that your key applications support the new silicon before buying.
For power users and gamers
Intel’s Series 3 and AMD’s Ryzen AI 400 bring enough CPU and GPU muscle to run demanding games and creative suites without leaning on a discrete GPU for AI features. The integrated Arc graphics in Intel’s new chips and Radeon 800M graphics in AMD’s mean you may not need to sacrifice battery life for on-device inference. However, raw TOPS numbers aren’t the whole story. Thermal headroom and memory bandwidth will determine how well these machines handle sustained workloads, so wait for independent benchmarks if you push your system hard.
For IT administrators and enterprise buyers
The HP EliteBoard G1a represents a new class of endpoint: a full Windows 11 PC inside a keyboard that connects to any monitor, simplifying hot-desking and rapid provisioning. But novel form factors introduce new management challenges. You’ll need to validate that devices like this work with your endpoint management tools, firmware update mechanisms, and security stacks. On the positive side, on-device AI reduces the amount of sensitive data that needs to travel to the cloud, which can simplify compliance. Still, you’ll need to review how data loss prevention (DLP) policies and telemetry interact with local Copilot processing.
For developers and ISVs
An 80 TOPS NPU in a mid-range Snapdragon laptop opens the door to more ambitious local models and agentic features. However, the market now spans x86 and Arm, with at least three distinct NPU architectures (Intel’s, AMD’s, Qualcomm’s). Testing and optimizing across all of them will add complexity. Prioritize cross-platform frameworks and quantization techniques to avoid getting locked into one vendor’s toolchain.
How we got here: from cloud-only AI to mandatory NPUs
Microsoft introduced the Copilot+ PC concept in early 2024, setting a 40 TOPS NPU threshold for advanced Windows AI features like Recall, live captions, and Studio Effects. At the time, only Qualcomm’s Snapdragon X Elite delivered the necessary performance, giving Windows on Arm a brief exclusive. Intel and AMD scrambled to catch up, shipping “AI-enhanced” chips with weaker NPUs throughout 2025 and promising future silicon. CES 2026 closes that gap. Intel’s 18A manufacturing milestone, AMD’s XDNA 2 architecture, and Qualcomm’s decision to push 80 TOPS into a mainstream X2 Plus SKU all signal that the industry now views on-device AI as a baseline platform capability, not a premium differentiator. The OEM designs at the show prove that PC makers are ready to compete on sustained AI performance, thermals, and form factor innovation rather than just spec-sheet TOPS.
What to do now: smart buying and deployment advice
If you’re planning to buy a new Windows laptop in 2026, here’s a practical playbook:
- Look past the TOPS label. A high peak TOPS number means little if the device throttles after two minutes. Insist on reviews that measure sustained NPU throughput, battery drain under AI workloads, and real-world latency in Copilot features.
- Match the silicon to your primary task. Creators and gamers who need raw compute should lean toward Intel Series 3 or AMD Ryzen AI 400 laptops with robust cooling. Road warriors who prize battery life should consider Snapdragon X2 Plus Arm devices, but test that your must-have x86 apps run without hiccups through emulation.
- Pilot before you roll out. If the HP EliteBoard G1a or similar keyboard PCs tempt your IT department, start with a small-scale trial. Validate manageability, driver update cadence, and ergonomic comfort before committing to a fleet.
- Check enterprise support contracts. Confirm Windows 11 Pro licensing, firmware update longevity, and integration with endpoint management platforms like Microsoft Intune. HP’s EliteBoard, for example, pitches modular repairability—but only an enterprise service agreement guarantees part availability years from now.
- Wait for independent benchmarks. CES demos are made to impress. Third-party testing will reveal whether these chips throttle under extended loads and how much AI actually improves daily tasks like video editing or spreadsheet work.
Outlook: what to watch over the next 12 months
The hardware foundation is now in place, but the software story is still being written. Copilot+ features will roll out gradually through Windows 11 updates, and app developers need time to integrate with the various NPU SDKs. Expect a wave of “AI-accelerated” stickers on laptops that may not all deliver meaningful gains. The real test will come when side-by-side comparisons show whether an 80 TOPS NPU makes a noticeable difference in common workflows versus a 50 TOPS unit. Also keep an eye on memory pricing: these processors demand faster, often soldered RAM to feed the NPU, which could keep entry prices high. If you can hold off, waiting for back-to-school or holiday 2026 refresh cycles will bring more choice and likely sharper prices. If you must buy now, prioritize a Copilot+ certified machine with a proven cooling design and a vendor that commits to driver support for at least three years.