Ubilink Supercomputing is built on the NVIDIA SuperPOD architecture, powered by 128 H100 GPU servers (1,024 GPUs) delivering up to 45.82 PetaFlops of performance. This unparalleled capability accelerates even the most demanding AI and HPC workloads with exceptional speed and efficiency.
Our platform integrates a high-performance AI-HPC parallel file system with NVIDIA GPUDirect RDMA, dramatically enhancing I/O throughput and enabling ultra-low-latency data access for large-scale AI training.
Moving forward, Ubilink will continue to expand its value-added computing ecosystem—offering a truly one-stop, high-performance AI and HPC solution for enterprises building the future of intelligence.
Compute Service Pricing
Ubilink offers two primary billing models for computing services:
- Pay-as-you-go (On-demand): Usage-based hourly billing
- Committed Term Rental: Customized discounted plans based on duration and quantity.
Please contact our sales team for tailored pricing.
Product Pricing
US$4
Per GPU-Hour
- Computing Specifications
- GPU：NVIDIA H100 (SXM5) 80GB × 1
- CPU：8 vCore
- RAM：192 GB DDR5-4800
- High-Speed Storage：125GB
- Supported Platforms: CloudInfra / Slurm