At the 2025 Heterogeneous Computing Architecture Ecosystem (HEC) Artificial Intelligence Innovation Conference, Sugon officially launched its groundbreaking domestically-developed 10,000-GPU AI supercomputing cluster—the scaleX 10,000-card system. Designed as a large-scale intelligent computing infrastructure for handling complex computational tasks, this system represents multiple technological innovations, with some advancements even surpassing the projected capabilities of similar overseas products expected by 2027.

The scaleX supercluster offers four key competitive edges:

• World's First 640-GPU Single-Cabinet Super Node: Achieving unprecedented density in AI computing hardware.
• Self-Developed Native RDMA High-Speed Networking: Enabling ultra-low-latency data transmission critical for AI workloads.
• Deeply Optimized Storage-Computing-Transmission Integration: Tightly coupling system components for maximum efficiency.
• Digital Twin Supercluster with Intelligent Scheduling: Real-time simulation and optimization of computing resources.

The system demonstrates remarkable versatility by supporting multi-brand accelerator cards and mainstream computing ecosystems. It is compatible with over 400 leading large language models and can be deployed across various application scenarios, from scientific research to industrial applications.