Russian
Giga Computing has launched GAIFA (GIGABYTE AI Factory Accelerator), its first in-house AI computing hub, located in New Taipei City, Taiwan. This six-rack, direct liquid-cooled facility is purpose-built to validate rack-scale AI infrastructure jointly with customers prior to official shipment and deployment. The first operational phase supports a 700 kW power budget, with additional capacity scheduled for future expansion based on business and operational demands. The site serves as an integrated verification platform to test and optimize full-stack compute performance, power delivery, and thermal management systems.
Rack-Scale Infrastructure and Compute Topologies
GAIFA deploys diverse accelerator platforms side-by-side to validate and benchmark end-to-end AI training, inference, and visualization workflows. Its core cluster consists of two liquid-cooled racks equipped with NVIDIA GB300 NVL72 systems, which integrate advanced GPUs, CPUs, high-speed interconnects and east-west networking. These configurations are optimized for trillion-parameter LLM training and complex agentic AI workloads.
High-density AI acceleration is delivered via one rack of G4L4-SD3-LAX7 direct liquid-cooled servers powered by NVIDIA HGX B300. Professional visual computing, generative AI inference and creative workloads run on XL44-SX2-AAS1 nodes equipped with NVIDIA RTX PRO 6000 GPUs. Meanwhile, liquid-cooled B683-Z80-LAS1 servers — a 6U ten-socket blade platform based on AMD EPYC processors — support dense host computing and virtualization tasks.
Full-Stack Validation and Proof of Concept Services
Building on Giga Computing’s existing proof-of-concept offerings, the new facility provides customers with a dedicated environment to comprehensively verify compute performance, rack power density, liquid cooling thermal efficiency, network stability and cross-component system compatibility before large-scale deployment. “GAIFA is more than Giga Computing’s internal AI computing center; it is a real-world testbed for co-validating enterprise-grade AI infrastructure with our customers,” stated Daniel Hou, General Manager at Giga Computing. The facility supports full-lifecycle validation up to L12 deployment standards, covering global service delivery and cluster-level operational verification.
Socket SP7 and SP8 Platform Roadmap
In tandem with the GAIFA launch, Giga Computing has expanded its server lineup with two new SP7-based platforms. The R165-DG2-CS1 is a 1U single-socket rack server with closed-loop liquid cooling, supporting up to 256-core configurations. The B685-D80-LS1 is a 6U ten-node blade server featuring direct liquid cooling, tailored for high-density HPC scenarios. The product release timeline remains consistent with previous announcements: SP7-based servers will launch alongside AMD’s platform release in November 2026, while next-generation SP8 systems are scheduled to debut in March 2027. Full technical specifications and operational details of the GAIFA facility will be released upon project completion.
Beijing Qianxing Jietong Technology Co., Ltd.
Sandy Yang/Global Strategy Director
WhatsApp / WeChat: +86 13426366826
Email: yangyd@qianxingdata.com
Website: www.qianxingdata.com/www.storagesserver.com
Business Focus:
ICT Product Distribution/System Integration & Services/Infrastructure Solutions
With 20+ years of IT distribution experience, we partner with leading global brands to deliver reliable products and professional services.
“Using Technology to Build an Intelligent World”Your Trusted ICT Product Service Provider!
Sandy Yang/Global Strategy Director
WhatsApp / WeChat: +86 13426366826
Email: yangyd@qianxingdata.com
Website: www.qianxingdata.com/www.storagesserver.com
Business Focus:
ICT Product Distribution/System Integration & Services/Infrastructure Solutions
With 20+ years of IT distribution experience, we partner with leading global brands to deliver reliable products and professional services.
“Using Technology to Build an Intelligent World”Your Trusted ICT Product Service Provider!