AI & Compute Infrastructure
AI Servers & GPU Systems
Rack-mounted GPU and accelerator servers for AI training, inference, high-performance computing and video analytics.
Product Series
Products
8-GPU AI Training Server (6U)
High-density air-cooled training platform for large models and HPC.
Typical configuration
- 2× latest-generation x86 server CPUs
- 8–10 double-width GPUs (up to 600 W each)
- Up to 4 TB DDR5
- PCIe 5.0
- Redundant power
Configuration confirmed per project
8-GPU AI Training Server (4U)
Compact dense training and HPC server.
Typical configuration
- 2× x86 server CPUs
- 8 double-width GPUs
- Up to 3–4 TB DDR5
- NVMe storage
- Up to 13 PCIe 5.0 x16 slots
Configuration confirmed per project
8-GPU Inference Server (4U/5U)
Mixed inference, rendering and video-processing workloads.
Typical configuration
- x86 server platform
- 8 double-width GPUs
- 9 PCIe 5.0 x16 slots
Configuration confirmed per project
Entry GPU Server (4U, 2-GPU)
Cost-effective platform for light AI, vision and video analytics.
Typical configuration
- 2× server CPUs
- 2 double-width GPUs
- Up to 1 TB DDR5
Configuration confirmed per project
AI Accelerator Supernode (10U)
Large-model training and inference on alternative (non-GPU) accelerator architectures.
Typical configuration
- 8 AI accelerators
- 4 server CPUs
- Up to 4 TB DDR5
- 400GE fabric
- Hybrid air/liquid cooling
Configuration confirmed per project
Domestic-Platform GPU Server
8-GPU server on China-made CPU platforms for localisation requirements.
Configuration confirmed per project
General-Purpose Compute Servers (2U)
Dual-socket rack servers for virtualisation, databases, MES/ERP and edge data processing.
Configuration confirmed per project
Edge AI Computing Units
Compact AI compute for machine vision and robot fleets at site level (details on request).
Configuration confirmed per project
Typical applications
LLM training and fine-tuningAI inferenceMachine visionSimulation & digital twinsVideo analytics