Acer Announces 'Veriton GN100', an Ultra-Compact AI Workstation for On-Premise Desk Environments

Acer Japan Corp. (Headquarters: Shinjuku-ku, Tokyo, President: Bob Sen) announces the launch of the 'Acer Veriton GN100', an ultra-compact AI workstation that supports AI development and the execution of large-scale models, available starting today, May 21, 2026, for corporate customers and on the official Acer online store.

This product is a new alternative for on-premise AI environments, packing high AI computing performance into a compact chassis that can be placed on a desk.

Start your on-premise AI environment right on your desk.

The Acer Veriton GN100 is a compact AI workstation based on the NVIDIA DGX Spark platform, designed to complete AI development and large-scale model execution internally. Powered by the NVIDIA GB10 Grace Blackwell Superchip, it achieves high AI computing performance of up to 1 petaFLOP. By running AI processes locally that previously relied on the cloud, it not only reduces cloud usage fees and data transfer costs but also enables secure operations where sensitive and research data never leaves the premises. Furthermore, by managing development, verification, and operation consistently in-house, it accelerates development speed and supports continuous technology accumulation, making it a new AI infrastructure choice that evolves enterprise AI utilization into something more realistic and sustainable.

Make large-scale AI workloads more efficient.

High-speed processing of large AI workloads: With the NVIDIA GB10 Superchip designed specifically for AI processing, 128GB of coherent unified system memory, and a high-speed 4TB SSD, it stably and smoothly executes AI development, data analysis, and model training handling large volumes of data. It reduces processing times in high-load workloads such as autonomously self-evolving AI agents, large language models, generative AI, image analysis, simulation, and scientific computing. This enables safe autonomous agents to run in local environments, accelerating the development cycle and increasing the number of verification iterations, thereby contributing to improved productivity in R&D and product development.

A high-performance AI foundation realized in a space-saving design.

Despite its compact size of approximately 150mm square and lightweight design of around 1.2kg, it offers processing performance comparable to an AI server. It can be installed in limited spaces such as office desks or laboratories, allowing users to build an AI environment without the need for large-scale server rooms. It also supports stacking connections of up to two units and is compatible with Kensington locks in consideration of physical security, providing an easy-to-deploy AI foundation that combines space-saving, high-performance, and secure features.

A next-generation AI platform with excellent scalability.

It supports high-speed network connections via the NVIDIA ConnectX-7, and by connecting multiple GN100 units, it can handle even larger AI workloads. A two-unit configuration supports inference for models up to 405 billion parameters, while a four-unit configuration using a 200 GbE switch with RoCE support can handle local inference for state-of-the-art AI models up to 700 billion parameters, communication-intensive workloads, and local AI factory operations. With low-latency and high-speed data transfer, it realizes a flexible foundation capable of handling large-scale AI processing, R&D, and future AI project expansions.

Supporting flexible deployment and operation with versatile interfaces.

For external interfaces, in addition to the NVIDIA ConnectX-7 NIC, it is equipped with three USB 3.2 ports (Type-C, Gen 2x2, 20Gbps), one USB 3.2 port (Type-C, Gen 2x2, 20Gbps, PD support), one HDMI 2.1 output port, and one Ethernet (RJ-45) port. Furthermore, it supports Wi-Fi 7, 10GbE, wired LAN, and Bluetooth 5.4, supporting high-speed data transfer, connection with external devices, and flexible integration into network environments. It easily accommodates equipment configurations suited to development and operational environments, offering the scalability and connectivity required for corporate use.

Smoother workflows from AI development to operation.

Adopting the NVIDIA DGX OS, it provides a stable Linux environment optimized for AI and machine learning, allowing users to complete the development, verification, and execution of AI agents and models locally utilizing development tools like PyTorch, Jupyter, and Ollama. Additionally, leveraging the NVIDIA AI software stack, including NVIDIA NemoClaw and NVIDIA OpenShell, it allows always-on agents to be securely deployed in a local environment.