Giga Computing, a GIGABYTE subsidiary, showcased its data center portfolio and NVIDIA Rubin platforms at NVIDIA GTC 2026, emphasizing scalable AI solutions for enterprises. The company introduced GIGABYTE AI TOP ATOM, a desktop supercomputer with 1 petaFLOPS performance via NVIDIA GB10 Grace Blackwell Superchips, and the W775-V10-L01 deskside system offering 20 petaFLOPS AI FP4 capabilities. These systems support trillion-parameter models using NVIDIA GB300 Grace Blackwell Ultra chips, targeting research institutions and enterprises.

GIGAPOD, GIGABYTE’s rack-scale AI/HPC solution, integrates NVIDIA Vera CPUs, Rubin GPUs, and NVLink 6 interconnects for exascale performance. Liquid-cooled clusters and air-cooled variants, paired with nVent CX121 CDU cooling systems, optimize efficiency. The GIGABYTE POD Manager software centralizes monitoring and orchestration, streamlining AI workload deployment across supercomputing infrastructures.

The Giga Computing AI Factory Accelerator (GAIFA) in Taiwan will serve as a hub for building AI factories, leveraging NVIDIA Quantum-X800 networking, Spectrum-X Ethernet, and AI Enterprise software. This facility aims to accelerate AI factory implementation globally, combining GIGABYTE’s infrastructure with NVIDIA’s hardware and frameworks. Notably, GIGABYTE’s XN24-VC0 server, selected by RIKEN Center, uses NVIDIA MGX architecture for scientific AI and quantum computing.

GIGABYTE also highlighted NemoClaw, an open-source stack for secure, local AI agents, and RTX PRO 6000 Blackwell GPUs for AI factories. With booth #1413 at NVIDIA GTC, the company positions itself as a leader in edge-to-cloud AI infrastructure, blending NVIDIA’s Vera CPUs and Rubin GPUs with advanced cooling and management tools to meet rising AI demands.