ASUS Ascent QN10: First Snapdragon X2 Elite Mini PC Hits 80 TOPS

The ASUS Ascent QN10 just made its debut at Computex 2026 on June 2, and it earns a spec-sheet headline that matters: it is the world's first mini PC built on Qualcomm's Snapdragon X2 Elite. For those of us who track silicon as closely as we track shipping firmware, this is the moment the X2 Elite leaves the laptop chassis and lands in a tiny, fanned-and-finned 0.7-liter box you can tuck behind a monitor. Let me walk through the numbers, because they tell the story better than any marketing slide.

Inside the ASUS Ascent QN10: Snapdragon X2 Elite silicon

The brain here is the Snapdragon X2 Elite, specifically model X2E-88-100. It is an 18-core Armv9 Oryon design split into two tiers: 12 Prime cores clocking up to 4.7 GHz and 6 Performance cores reaching up to 3.4 GHz. That asymmetric layout is built for the way real workloads behave. Bursty, latency-sensitive tasks like compiling, opening a heavy IDE, or spinning up a container land on the high-clock Prime cores, while sustained background jobs spread across the Performance cores to keep efficiency and thermals in check.

Feeding those cores is up to 32GB of LPDDR5X-9600 memory. The 9600 MT/s figure is the part worth noting. On an Arm system-on-chip, the CPU, GPU, and NPU all share that memory pool, so high bandwidth directly translates into faster model loads and snappier multitasking. Storage is generous for the form factor: up to 4TB across two M.2 2280 PCIe Gen4 slots, which means you can run a fast boot drive and a roomy data drive without external enclosures.

What 80 TOPS means for local AI

The piece that makes this a genuine Copilot+ PC is the Hexagon NPU, rated at 80 TOPS using INT8 math. TOPS stands for trillions of operations per second, and INT8 refers to 8-bit integer precision, the format most on-device inference runs in. So 80 TOPS means the chip can perform roughly 80 trillion of those low-precision multiply-accumulate operations every second, dedicated entirely to AI work.

Here is why that number is more than a bragging point. A dedicated NPU runs inference without waking the CPU or GPU, which keeps power draw and heat low while leaving general compute free. In practical terms, 80 TOPS comfortably clears the 40-plus TOPS bar Microsoft set for Copilot+ experiences, so features like live transcription, real-time translation, image generation, and on-device assistants run locally. For developers, that means small language models, speech pipelines, and vision models can execute on the box itself: no cloud round-trip, no per-call latency, and your data never leaves the desk.

A connectivity loadout built for makers and integrators

ASUS did not skimp on I/O, which is exactly what you want from a machine aimed at AI developers and systems integrators. The QN10 carries HDMI 2.1, three USB4 ports at 40Gbps each, 2.5GbE wired networking, WiFi 7, and Bluetooth 5.4. That USB4 trio alone is a strong analysis point: 40Gbps lanes give you fast external NVMe, high-resolution capture, or eGPU-class peripherals. The system can drive up to four 4K displays, so a multi-monitor development or monitoring wall is on the table out of the box.

All of this lives in a chassis measuring roughly 130 x 130 x 40 mm and weighing about 720 grams. That 0.7-liter volume is what makes the QN10 compelling for deployment: home-office desks, digital-signage racks, kiosks, and edge nodes where a tower simply will not fit. It runs Windows 11 as a Copilot+ PC, so the familiar tooling, drivers, and software stack come along for the ride.

The takeaway for developers and home offices

ASUS has not announced pricing yet, so I will hold off on value judgments until those figures land. What is confirmed is the recipe: an 18-core Oryon CPU, an 80 TOPS NPU, fast LPDDR5X, ample Gen4 storage, and modern I/O in a palm-sized enclosure. For makers, AI developers, and integrators who want a compact local-AI box that sips power while taking inference off the cloud, the Ascent QN10 is a genuinely exciting first-of-its-kind machine. It is a great sign of where efficient Arm computing is heading.

Sources: CNX Software, June 2, 2026; Qualcomm, June 2026; ASUS, June 2026