


Gigabyte AI TOP Atom / 4TB Storage + 128GB & Unified System Memory
The GIGABYTE AI TOP Atom (4TB Storage + 128GB Unified System Memory) is a compact, purpose-built local AI workstation designed for developers and creators who want to prototype, fine-tune, and run modern AI models on-device without relying on the cloud. Built around the NVIDIA GB10 Grace Blackwell Superchip, it combines a 20-core Arm CPU (10× Cortex-X925 + 10× Cortex-A725) with NVIDIA Blackwell architecture and 128GB LPDDR5x coherent unified system memory, enabling large-model workflows and fast CPU↔GPU data sharing for AI development. It comes with a 4TB PCIe Gen5 NVMe SSD for large datasets and projects, and runs on NVIDIA DGX™ OS (Ubuntu Linux) for a ready-to-use AI software environment.
Key Features
-
NVIDIA GB10 Grace Blackwell Superchip built for local AI workloads (from prototyping and fine-tuning to inference and data science)
-
20-core Arm CPU: 10× Cortex-X925 + 10× Cortex-A725 for strong multi-threaded compute and data prep tasks
-
128GB LPDDR5x coherent unified system memory for large models and high-throughput AI workflows (shared memory architecture)
-
4TB PCIe Gen5 NVMe SSD for high-capacity, high-speed local dataset/project storage
-
Supports up to ~200B parameter models (positioned for large-model local development/inference)
-
NVIDIA DGX™ OS + Ubuntu Linux for an AI-ready software stack out of the box
Product Information
Product Information
Shipping & Returns
Shipping & Returns
Description
The GIGABYTE AI TOP Atom (4TB Storage + 128GB Unified System Memory) is a compact, purpose-built local AI workstation designed for developers and creators who want to prototype, fine-tune, and run modern AI models on-device without relying on the cloud. Built around the NVIDIA GB10 Grace Blackwell Superchip, it combines a 20-core Arm CPU (10× Cortex-X925 + 10× Cortex-A725) with NVIDIA Blackwell architecture and 128GB LPDDR5x coherent unified system memory, enabling large-model workflows and fast CPU↔GPU data sharing for AI development. It comes with a 4TB PCIe Gen5 NVMe SSD for large datasets and projects, and runs on NVIDIA DGX™ OS (Ubuntu Linux) for a ready-to-use AI software environment.
Key Features
-
NVIDIA GB10 Grace Blackwell Superchip built for local AI workloads (from prototyping and fine-tuning to inference and data science)
-
20-core Arm CPU: 10× Cortex-X925 + 10× Cortex-A725 for strong multi-threaded compute and data prep tasks
-
128GB LPDDR5x coherent unified system memory for large models and high-throughput AI workflows (shared memory architecture)
-
4TB PCIe Gen5 NVMe SSD for high-capacity, high-speed local dataset/project storage
-
Supports up to ~200B parameter models (positioned for large-model local development/inference)
-
NVIDIA DGX™ OS + Ubuntu Linux for an AI-ready software stack out of the box



















