All Data Can Be AI Value
AI factories are a type of IT infrastructure that transforms organizational data into AI inventions and propels industry leaders ahead of the competition. GIGABYTE Technology, an end-to-end data center infrastructure and AI solution provider, can help you establish your AI factory. Our time-tested, world-renowned hardware and software portfolio can turn data pipelines into round-the-clock generators of smart value. GIGABYTE AI Factory Accelerator (GAIFA)—our self-designed, self-built, Taiwan-based AI factory—can expedite testing, validation, and deployment, putting you on track to enjoy unprecedented AI success.
Agentic AI & Digital Workforce
AI agents coordinating a framework of AI models trained on internal data can achieve business goals with minimal human intervention.
Physical AI in Automation & Robotics
Predictive maintenance, quality control, and robotic helpers capable of true autonomy expand the benefits of AI to real-world applications.
Digital Twins & HPC-driven Simulations
Enhance cost-efficiency and shorten time-to-market by running insightful, high-precision tests during the research and development process.
Product Design & Content Creation
Generative AI cognizant of an organization's strategy and vision can help brainstorm new products as well as create marketing content.
.
Develop Your AI Foundation with GIGAPOD
At the core of the AI factory, scalable supercomputing clusters convert big data into AI tokens at breakneck speed by utilizing GPU-centric configurations designed for deep learning and other AI training methodologies. GIGABYTE's GIGAPOD combines multiple arrays of state-of-the-art GPUs through blazing-fast interconnect technology to form a cohesive unit that serves as the building block of modern AI. Clients can not only opt for AMD Instinct™, Intel® Gaudi®, or NVIDIA HGX™ GPU modules, but they can also choose between air cooling and liquid cooling to strike a perfect balance between investment and oomph. GIGAPOD is rounded out with centralized control for infrastructure oversight, the middleware platform G-REX for power and cooling management, proprietary GPM software for DCIM and workload orchestration, as well as pre-integrated single-rack GIGAPOD Rack Scale systems for faster deployment and activation.
Air-CooledLiquid-Cooled
*Compute Racks + Management Rack
| Supported GPUs | Server Height | Power Consumption per Rack | No. of Racks per SU* | PDU per Rack |
|---|---|---|---|---|
| NVIDIA HGX™ B300 | 8U | 70 kW / 66 kW | 8 + 2 / 8 + 1 | 8 x 63A / 4 x 63A |
| NVIDIA HGX™ B200 | 8U / 8OU | 70 kW / 54 kW | 8 + 1 | 8 x 63A / 4 x 63A |
| NVIDIA HGX™ H200 | 8U | 58 kW | 8 + 1 | 4 x 63A |
| NVIDIA RTX PRO™ 6000 Blackwell Edition | 4U | 80 kW | 4 + 1 | 4 x 63A |
| AMD Instinct™ MI350 Series | 8U | 70 kW | 8 + 1 | 4 x 63A |
| AMD Instinct™ MI300 Series | 5U / 8U | 50kW - 100kW | 4 + 1 / 8 + 1 | 2 x 100A / 4 x 63A |
| Intel® Gaudi® 3 | 8U | 62kW | 8 + 1 | 4 x 63A |
.
Achieve Reasoning AI with NVIDIA Vera Rubin
To successfully deploy AI at scale, your computing needs to keep pace with your AI workloads. GIGABYTE collaborates with NVIDIA to present Vera Rubin NVL72 and Rubin NVL8—state-of-the-art AI platforms purpose-built to accelerate intelligent reasoning, autonomous decision-making, and scalable AI infrastructure for agentic and reasoning AI.
The rack-level Vera Rubin NVL72 unifies 72 Rubin GPUs, 36 Vera CPUs, ConnectX®-9 SuperNIC™s, BlueField®-4 DPUs, and NVIDIA NVLink™ 6 switch to turbocharge AI projects. The server-size Rubin NVL8 provides 400 petaFLOPS of inference performance and 160TB/s HBM to deliver super-intelligence. Both are best-in-class choices for next-gen AI factories.
The rack-level Vera Rubin NVL72 unifies 72 Rubin GPUs, 36 Vera CPUs, ConnectX®-9 SuperNIC™s, BlueField®-4 DPUs, and NVIDIA NVLink™ 6 switch to turbocharge AI projects. The server-size Rubin NVL8 provides 400 petaFLOPS of inference performance and 160TB/s HBM to deliver super-intelligence. Both are best-in-class choices for next-gen AI factories.
1 Management Switches
- 2 x Out-of-Band (OOB) management switches
2 2 x 3U 110 kW Power Shelves
3 10 x Compute Trays
- 1U XN16-CB0-L01, each with 2 x Vera CPUs & 4 x Rubin GPUs
4 9 x NVIDIA NVLink™ Switches
- 1U NVLink Switch Tray
- Sixth-generation NVLink with up to 3.6 TB/s GPU-GPU interconnect
5 8 x Compute Trays
- 1U XN16-CB0-L01, each with 2 x Vera CPUs & 4 x Rubin GPUs
6 2 x 3U 110 kW Power Shelves
7 Compatible with in-row CDU
Inference at Scale with NVIDIA GB300 NVL72
The next step after developing your AI models is to build applications and tools based on said models and deploy them throughout your organization, where hundreds or even thousands of users might access them at any given moment. You need a superscale inference platform capable of handling a multitude of requests simultaneously, so that your AI success can supercharge productivity. NVIDIA GB300 NVL72 features a fully liquid-cooled, rack-scale design that unifies 72 NVIDIA Blackwell Ultra GPUs and 36 Arm®-based NVIDIA Grace™ CPUs in a single platform optimized for test-time scaling inference. AI factories powered with the GB300 NVL72 using NVIDIA Quantum-X800 InfiniBand or Spectrum™-X Ethernet paired with NVIDIA ConnectX®-8 SuperNIC™ provide a 50x higher output for reasoning model inference compared to the NVIDIA Hopper™ platform, making them the undisputed leader in AI inference.
NVIDIA Rubin NVL8
| GPU | 8x NVIDIA Rubin GPUs |
| Total GPU Memory | Bandwidth | 2.3 TB | 160 TB/s |
| CPU | 2x Intel® Xeon® 6 processors |
| NVIDIA NVLink Switch System | 4x |
| NVIDIA NVLink Bandwidth | 28.8 TB/s total bandwidth |
| Networking | 8x OSFP ports serving 8x single-port NVIDIA ConnectX®-9 VPI - up to 800 Gb/s NVIDIA InfiniBand and Ethernet 2x 400G QSP112 NVIDIA BlueField®-4 DPUs - up to 800 Gb/s NVIDIA InfiniBand and Ethernet |
XL44-SX2-AAS1 with RTX PRO™ 6000 Blackwell Server Edition GPUs
- NVIDIA RTX PRO™ server with ConnectX®-8 SuperNIC switch
- Configured with 8 x NVIDIA RTX PRO™ 6000 Blackwell Server Edition GPUs
- Configured with 1 x NVIDIA® BlueField®-3 DPU
- Onboard 400Gb/s InfiniBand/Ethernet QSFP ports with PCIe Gen6 switching for peak GPU-to-GPU performance
- Dual Intel® Xeon® 6700/6500-Series Processors
- 8-Channel DDR5 RDIMM/MRDIMM, 32 x DIMMs
- 2 x 10Gb/s LAN ports via Intel® X710-AT2
- 8 x 2.5" Gen5 NVMe hot-swap bays
- 2 x M.2 slots with PCIe Gen4 x2 interface
- 3+1 3200W 80 PLUS Titanium redundant power supplies
Reimagine Your Infrastructure with Experts
One-Stop AI Infrastructure
GIGABYTE brings decades of L12 expertise to data center establishment, integrating hardware, software, and architecting solutions. If you're looking to erect an AI data center, GIGABYTE can simplify rollout, boost efficiency, and ensure performance with ease. Our services include:
- Project consultation and evaluation
- Site planning and floorplan design
- Rack and server design, configuration, and installation
- Cabling design and installation
- Power and thermal assessment
- Data center deployment and validation
GAIFA: GIGABYTE AI Factory Accelerator
GIGABYTE AI Factory Accelerator (GAIFA) is our self-designed, self-built AI factory showcasing the ability to realize centralized computing at scale and optimize for peak performance. The pinnacle of engineering know-how, GAIFA delivers full-stack solutions powered by NVIDIA's AI infrastructure, NVIDIA Quantum-X800 InfiniBand and Spectrum-X Ethernet networking, and GIGABYTE POD Manager (GPM) software paired with NVIDIA AI Enterprise. The Taiwan-based facility will expedite testing, validation, and deployment for our esteemed customers, enabling fast and seamless AI implementation.
Contact Us
GADU: GIGABYTE Accelerated Deployment Unit
The new GIGABYTE Accelerated Deployment Unit (GADU) redefines agile, customized, scalable AI integration. By packing IT nodes into standalone containers and supporting them with power and cooling modules, GADU can achieve rapid-deploy, plug and play AI infrastructure with minimal footprint in variable environments. Think of it as AI factory in a box, made to order and ready to go. See if GADU can spur a breakthrough in your AI strategy!
Contact Us
Accelerated Rollout
Prefabricated modules drastically reduce the time it takes to establish a data center, enabling clients to engage in AI transformation up to 400% faster.
Best-fit Architecture
Modular data centers can be tailor-made according to client specifications, with the option to quickly scale up or out by adding on new nodes.
Flexible Deployment
Insulating IT infrastructure in a safe, self-contained environment paves the way for proactive expansion in remote and rugged areas on the network edge.
Performance Perfected for AI Leadership
Advanced Cooling for Green Computing
Direct liquid cooling (DLC) and immersion cooling are revolutionary ways of unleashing the full potential of cutting-edge processors while improving power usage effectiveness (PUE) and reducing carbon emission. To help your AI factory achieve sustainable results, GIGABYTE extends its line of advanced cooling products from field-proven systems and racks to the overarching infrastructure.
We can help set up piping and cooling loops in a liquid-cooled environment or revitalize conventional air-cooled data centers with tailor-made solutions. Enhanced thermal and power efficiency equate to better return on investment for the environmentally conscious growth engines of the future.
Software Stack for AI & HPC Environment
GIGABYTE's offerings go beyond hardware to encompass full stack support. We provide a software package that gives users control over infrastructure and creates the perfect environment for task orchestration. Professionals who need ready-to-go AI & HPC total solutions choose GIGABYTE.
At the infrastructure level, GIGABYTE POD Manager (GPM) serves as a single point of access to all nodes, as well as the central overseer of the resource pool. GPM Workload Management leverages AIOps to streamline workflow, making it highly suited to AI development and deployment. Remote management, data security, and many more features are part and parcel of our solution stack. With this addition, GIGABYTE is pleased to present our total AI factory solution.
GIGABYTE AI Architecture with NVIDIA AI Enterprise
GIGABYTE's full-stack software suite GPM incorporates NVIDIA AI Enterprise (NVAIE), a set of cloud-native software tools that provide an optimized framework for seamless AI development, to vastly improve your AI factory efficiency. The inclusion of Run:ai and NVIDIA Mission Control enables automated orchestration and infrastructure management, while virtual GPU (vGPU) technology solidifies the foundation of a multi-tenant architecture, guaranteeing maximum output from GPU clusters like the GIGAPOD. GIGABYTE's software and hardware integration transforms complex hardware monitoring into highly efficient software-defined services that put enterprises in complete control of the entire AI lifecycle, from model training to deployment and inference.
Revolutionize AI Factory with CXL Memory Pooling
As AI models grow more complex, AI factories need to handle an enormous flow of data with efficiency and speed. CXL (Compute Express Link) memory pooling energizes AI infrastructure by optimizing resources, accelerating data movement, and supporting scalable growth. Bottlenecks are avoided by replacing a siloed model with an open-standard interconnect that allows CPUs, GPUs, and accelerators to tap into a shared pool.
GIGABYTE servers like R284-S91-AAJ2 support other servers in the cluster with CXL memory expansion modules that form a collective memory pool. Another option is G494-SB4-AAP2, which pairs PCIe Gen5 with CXL to extend memory capacity, enhance data exchange efficiency, and enable resource sharing.
Parallel File Systems & Flash Storage Solutions
To meet AI factories' massive data throughput requirements, GIGABYTE provides parallel file systems featuring options from WEKA and VAST Data.
Organizations can leverage WEKA's distributed architecture for ultra-low latency to unlock all-flash array (AFA) storage bandwidth and eliminate AI training bottlenecks. Alternatively, our solution with VAST Data offers a revolutionary unstructured data management architecture that ensures extreme scalability and reliability for petascale computing. By springboarding off of our expertise in cutting-edge technological integration, GIGABYTE can deliver AI storage solutions tailor-made for diverse application scenarios and computing demands.
Why The Best Choose GIGABYTE
Guaranteed Performance
GIGABYTE's AI factory solutions underwent rigorous testing to deliver unbeatable, unwavering performance in service of intensive workloads. The seamlessly compatible, extensively certified hardware and software will excel at any AI computing task.
Efficient Cooling
GIGABYTE provides air cooling, direct liquid cooling, and immersion cooling options meticulously designed to effectively manage heat at both the system and infrastructure level, ensuring peak output and improved sustainability for your AI factory.
Versatile Ecosystem
GIGABYTE's server lineup gives you a wide selection of AMD's and Intel's x86 chips, Ampere's and NVIDIA's ARM architecture, and specialized processing units from DPUs to NPUs, so you will always have the best setup for your AI mission.
Shorter TTM for AI Adoption
GIGABYTE takes pride in short time-to-market (TTM) solutions that address rapidly evolving client demands. GIGABYTE can customize product configurations to specific use cases, reducing cost, streamlining design, and enabling flexible rollout with minimal modifications.
Dynamic Scalability
GIGABYTE's AI factory solutions are built with future expansion in mind. Our turnkey AI infrastructure emphasizes interconnectivity and integrability so new accelerators, nodes, or clusters can be brought online in record time and begin generating value.
End-to-end Support
GIGABYTE works closely with industry partners to set a high bar for fast responses to customer requirements and timelines. GIGABYTE's service centers and expert teams across the globe are standing by to satisfy your AI factory support needs.
Our Ecosystem Partners
