ASUS unveils ultrasmall-form-factor PCs with a focus on AI performance at CES 2026
ASUS unveils compact AI-powered PCs at CES 2026, spanning enterprise, productivity, gaming and desktop-scale AI computing.
ASUS has introduced a new range of ultrasmall-form-factor PCs at CES 2026, positioning artificial intelligence performance, compact design and reliability as core priorities for its next generation of desktop systems. The announcement expands the company’s NUC and small-form desktop portfolio, targeting professional, enterprise, gaming and AI development use cases.
Table Of Content
The lineup includes the ASUS NUC 16 Pro, ExpertCenter PN55, ROG GR70 Gaming Mini PC and the ASUS Ascent GX10 desktop AI supercomputer. Each system is designed around a different workload profile, from AI-assisted productivity and enterprise deployment to high-end gaming and local AI model development. Despite their small physical footprints, all systems emphasise high compute density, modern connectivity standards and thermal efficiency.
Unveiled in Las Vegas, the new devices reflect ASUS’ broader strategy to push AI computing closer to the edge. Rather than relying solely on cloud infrastructure, the company is positioning compact desktops as capable platforms for local inference, development and secure workloads, particularly in environments where data control and latency matter.
Compact desktops built for AI-first workloads
At the centre of ASUS’ enterprise-focused announcements is the NUC 16 Pro Mini PC, designed for creators, developers and business users working with AI-enabled applications. The system is powered by up to an Intel Core Ultra X9 Series 3 processor with Intel Arc B390 graphics, delivering up to 180 platform TOPS of AI performance. According to ASUS, this represents nearly double the AI capability of its predecessor.
The NUC 16 Pro supports LPDDR5 and LPDDR5x memory, enabling up to 20 percent faster performance while reducing power consumption by as much as 50 percent. These gains are intended to support sustained workloads such as machine learning, analytics and intelligent applications without compromising efficiency. Dual 2.5G LAN ports provide stable wired connectivity, while WiFi 7 and Bluetooth 6.0 support high-speed wireless networking in dense environments.
A key feature of the system is ASUS AI SuperBuild, which provides tools for training and customising large language models locally. This approach allows AI agents to run on-device, reducing reliance on cloud services and helping organisations keep sensitive data within their own infrastructure. The NUC 16 Pro is housed in a 0.7-litre chassis and uses an intelligent dual-fan cooling system to manage thermals and noise under load.
The system is also positioned for demanding enterprise environments. ASUS states that the NUC 16 Pro has been tested beyond military-grade standards, including exposure to extreme temperatures, humidity, drops and vibration. This focus on durability aligns with the company’s emphasis on reliability for long-term deployment in offices, industrial settings and edge computing scenarios.
Copilot+ productivity and high-end gaming in small form factors
For productivity-focused users, ASUS introduced the ExpertCenter PN55 Mini PC, a Copilot+ PC built around the AMD Ryzen AI 400 Series processor. The system integrates Radeon 800M graphics and an NPU capable of delivering up to 55 TOPS, accelerating AI-driven workloads while maintaining energy efficiency and data security.
The PN55 includes a dedicated Copilot key for quick access to AI features, alongside a fingerprint sensor, built-in microphone and internal speaker to support secure authentication and voice-based interaction. Connectivity is a key focus, with dual 2.5G LAN ports, six USB ports, WiFi 7 and Bluetooth 5.4 designed to accommodate a wide range of peripherals and workflows. ASUS positions the PN55 as suitable for both consumers and businesses, with military-grade testing intended to ensure long-term reliability.
On the gaming side, ASUS revealed the ROG GR70 Gaming Mini PC, marking the first ROG-branded mini PC to feature an AMD Ryzen 9 processor. The system pairs the CPU with an NVIDIA GeForce RTX 5070 or RTX 5060 Laptop GPU, delivering what ASUS describes as desktop-class gaming performance in a compact enclosure.
The GR70 uses a triple-fan cooling system to manage heat during intensive gaming sessions, balancing thermal performance with low noise levels. Built-in WiFi 7 and Bluetooth 5.4 are designed to support low-latency online gaming and peripheral connectivity. The device targets users who want high-end gaming capability without the size and complexity of a traditional desktop tower.
Desktop-scale AI computing in a minimal footprint
Rounding out the announcements is the ASUS Ascent GX10, a compact desktop AI supercomputer designed for advanced AI workloads. Powered by the NVIDIA GB10 Grace Blackwell Superchip and equipped with 128GB of unified memory, the system delivers petaflop-scale AI performance in a form factor small enough for desktop use.
The Ascent GX10 is designed to support the rapid development of large generative models, running as a standalone system or scaling into a private cloud environment. Two units can be paired via NVIDIA ConnectX-7 networking to expand capacity, offering flexibility for research teams, developers and enterprises building local AI infrastructure.
Despite its size, ASUS positions the Ascent GX10 as a full-stack AI solution. It is accelerated by the NVIDIA AI software stack, enabling secure, high-performance local computing without immediate dependence on external data centres. This approach reflects growing demand for on-premise AI capabilities, particularly in sectors with strict data governance or latency requirements.
Collectively, the new lineup highlights ASUS’ intent to redefine what small-form desktops can deliver. By combining high-performance processors, dedicated AI acceleration and modern connectivity within compact designs, the company is addressing a market that increasingly values local intelligence, efficiency and reliability alongside raw compute power.