The Pulse

Gigabyte launches compact AI desktop with up to 128GB of memory

Gigabyte's AI TOP AERO combines Nvidia's RTX Spark platform with up to 128GB of unified memory in a desktop of approximately one liter. The Windows 11 machine is intended for local AI development, content creation and supported games.

AI.info Team · · fact-checked · against 5 sources ↓ · Published 12 Nov 2026

Gigabyte's new AI TOP AERO puts an Nvidia RTX Spark chip and up to 128GB of unified memory in a desktop with a chassis of approximately one liter. The Windows 11 machine is intended for local AI development, content creation and gaming, according to the company's announcement.

Gigabyte dates the launch to November 10, 2026. Regional configurations, prices and availability are to be covered in local announcements.

Hardware in a compact chassis

The desktop uses the Nvidia RTX Spark N1X Superchip, which combines a 20-core Nvidia Grace CPU and a Blackwell RTX GPU. Configurations offer up to 128GB of LPDDR5x unified system memory and up to a 4TB PCIe Gen5 solid-state drive.

The company puts its own cooling design inside the approximately one-liter chassis. Gigabyte says the cooling is intended to help sustain performance during extended local AI and professional workloads.

Connections include USB Type-C, 10GbE Ethernet and Wi-Fi 7 for peripherals and network access. The machine runs Windows 11 and is designed for work ranging from model development and content creation to everyday productivity and supported games.

Models and local networks

The CUDA platform lets developers run supported models locally, develop applications, fine-tune models and test prototypes. Supported Nvidia Nemotron models and CUDA-accelerated tools can be used for local inference, model testing and agentic AI applications.

Compatible versions of LM Studio and Ollama offer additional ways to set up a local AI environment. The desktop is aimed at users who want to test model responses, connect tools and build automated workflows on their own equipment.

A separate Nvidia tool, Personal AI Router, or PAIR, can direct individual inference requests to suitable devices on a local network. That setup requires compatible systems and supported models and inference engines; depending on the configuration, processing takes place on one of the connected devices.

In October, Nvidia said the official PAIR release added RTX Spark and llama.cpp support, direct model downloads from Hugging Face and configurable inference-engine settings. The free, open-source tool discovers and connects compatible devices and distributes AI workloads across available systems.

Creative tools and games

For creators, the desktop can work with compatible tools such as ComfyUI to build generative AI workflows. Nvidia Broadcast and RTX Video provide audio and video enhancement features in supported applications, covering work from generated content through post-production.

Nvidia's Broadcast 2.3 runs natively on RTX Spark and includes video effects and an offline file processor for applying effects to prerecorded material. The release also introduced controls that let local AI agents adjust settings and process files through the app.

For games, the desktop's Blackwell GPU supports Nvidia DLSS 5. Gigabyte says supported games can use its neural-rendering features for lighting and materials.