GIGABYTE has officially launched its next-generation personal AI supercomputer, the AI TOP ATOM, marking a new chapter in desktop-scale high-performance computing. Built to deliver supercomputer-level capability in a compact, energy-efficient form, the system is powered by NVIDIA’s Grace Blackwell GB10 Superchip and comes preloaded with the NVIDIA AI Software Stack for seamless AI development and deployment.
Supercomputer Power in Compact Form
The AI TOP ATOM transforms what users can expect from a desktop machine. Equipped with 128 GB of unified system memory and up to 4 TB of SSD storage, it delivers an astonishing FP4 AI performance of up to 1 PetaFLOP. This enables local execution of models with up to 200 billion parameters, bridging the gap between professional data centers and creative workstations.
Modular Scaling for Massive Models
For even greater power, users can connect two AI TOP ATOM units using the integrated NVIDIA ConnectX-7 networking card. This dual setup allows model training and inference for up to 405 billion parameters, making it suitable for large-scale research, simulation, and real-time generative AI workflows. Despite its strength, the chassis remains compact and whisper-quiet, ensuring maximum flexibility in labs, studios, or classrooms.
AI Development Simplified
Preloaded with the NVIDIA AI Software Stack, this system offers a full suite of tools, frameworks, and libraries optimized for generative AI. Developers can fine-tune large language models (LLMs), multimodal systems (LMMs), and machine learning pipelines directly on the desktop. Moreover, GIGABYTE’s exclusive AI TOP Utility introduces an intuitive interface for managing inference, deployment, and performance optimization—ideal for both professionals and learners entering the AI domain.
Empowering Research and Innovation
The AI TOP ATOM represents GIGABYTE’s commitment to democratizing AI computing. By combining scalability, efficiency, and accessibility, it empowers developers, institutions, and startups to experiment with next-gen AI without relying on costly cloud infrastructures. The system will be available worldwide starting October 15, 2025.
| Specification | Details |
|---|---|
| Processor | NVIDIA Grace Blackwell GB10 Superchip |
| AI Performance | Up to 1 PetaFLOP (FP4) |
| Memory | 128 GB Unified System Memory |
| Storage | Up to 4 TB SSD |
| Networking | NVIDIA ConnectX-7 (supports dual-unit link) |
| Model Capacity | Up to 200B parameters (single) / 405B (dual) |
| Software | NVIDIA AI Software Stack + GIGABYTE AI TOP Utility |
| Launch Date | October 15, 2025 |
