CES 2026: Intel, AMD and Qualcomm push AI PCs to 50-80 TOPS, and Intel takes aim at edge systems

CES 2026 settled one thing about the AI PC: the three major laptop silicon vendors are now competing on a headline number, NPU TOPS, and all three arrived in the same week. Intel, AMD and Qualcomm each announced new processors with dedicated neural processing units rated between 50 and 80 TOPS ^[1][2][3]. The more interesting developments for businesses are less about that number than about what surrounds it: laptop availability dates, a developer box built around large unified memory, and Intel's push into industrial edge systems.

What each vendor announced

Intel says more than 200 designs are planned across partners, with additional designs through the first half of 2026 ^[1]. AMD lists Acer, ASUS, Dell, HP, GIGABYTE and Lenovo as first-quarter system makers ^[2]. Qualcomm's chip pairs the NPU with a third-generation Oryon CPU of up to 10 cores and LPDDR5x memory, and the company calls it "the fastest NPU in a laptop" ^[3].

Why the headline number deserves caution

AMD's own release is candid about the limits: its footnote says TOPS values are maximums "in an optimal scenario" and may vary with system configuration, AI model and software version ^[2]. The vendor materials reviewed for this article do not provide a measured result for a business workload such as summarizing documents or transcribing meetings locally. A 60 TOPS part is not necessarily 20 percent faster than a 50 TOPS part on the software you run, and the 80 versus 50 gap says even less. Treat the figures as a minimum bar for supporting on-device features, not as a ranking.

The more practical detail: memory

AMD's Ryzen AI Halo is a developer mini-PC built on Ryzen AI Max+ processors, scheduled for the second quarter of 2026, with up to 128GB of unified memory and ROCm preinstalled. AMD says it can run models of up to 200 billion parameters locally, and its footnote specifies that this requires the 128GB configuration ^[2]. The takeaway is that local model capability is limited first by memory capacity. AMD's Ryzen AI Max+ 392 and 388 chips carry a 50 TOPS rating ^[2], so the platform's selling point is memory and graphics, not the NPU.

Intel's edge push

Intel also announced that Series 3 edge processors, available starting in the first quarter of 2026, are certified for embedded and industrial use, with extended temperature ranges, deterministic performance and 24x7 reliability. Target uses include robotics, smart cities, automation and healthcare ^[1]. Intel estimates up to 1.9x higher LLM performance, up to 2.3x better performance per watt per dollar on end-to-end video analytics, and up to 4.5x higher vision-language-action model throughput than Nvidia's Jetson Orin AGX 64GB ^[1]. These are Intel's estimates and have not been independently reproduced in the sources reviewed here.

What this means for businesses

Do not select laptops on TOPS. Pick two or three representative tasks, such as meeting transcription or local document search, run them on pilot hardware, and measure time and battery impact, while sizing memory for the largest local model you actually intend to run. For edge and industrial deployments, ask vendors for benchmarks on your workload rather than relying on comparisons against another vendor's flagship.