As AI adoption accelerates across industries, enterprises are moving beyond model development to large-scale AI deployment. The rapid growth of large language models (LLMs), intelligent agents, and AI-powered applications is driving unprecedented demand for high-performance, energy-efficient computing infrastructure.

However, heterogeneous AI environments remain a major challenge. Proprietary architectures, limited hardware compatibility, rising deployment costs, and increasing power consumption continue to hinder enterprise AI adoption.

To address these challenges, Gooxi G6A80A5 OAM AI Training & Inference Server delivers an open architecture, exceptional computing performance, intelligent cooling, and flexible scalability. Designed for modern AI workloads, it helps organizations simplify heterogeneous deployment while building a future-ready AI infrastructure.

Open Architecture for Greater Flexibility

Built on the OAM 2.0 open standard, the G6A80A5 supports mainstream AI accelerator modules from both domestic and international vendors. It natively supports MetaX C550 OAM 2.0 modules, offers seamless compatibility with SXM-based accelerators, and is designed to accommodate next-generation OAM modules, enabling true plug-and-play upgrades.

This open architecture gives enterprises greater freedom to select the best AI accelerators for their workloads while reducing deployment complexity, migration costs, and vendor lock-in.

Built for High-Performance AI Workloads

G6A80A5 supports up to two AMD EPYC™ 9005 Series processors, with up to 192 cores per CPU and a maximum 500W TDP. Combined with 24 DDR5-6400 memory slots and a 12-channel memory architecture, the platform delivers the bandwidth and throughput required for LLM training, AI inference, deep learning, real-time analytics, and other compute-intensive applications.

Intelligent Hybrid Cooling for Greater Efficiency

To meet the cooling demands of high-density AI computing, the system features 15 high-speed rear fans, providing up to 1200W cooling capacity per accelerator module in air-cooled configurations.

The integrated air- and liquid-cooling design enables seamless switching between cooling methods, ensuring compatibility with next-generation high-power OAM accelerators. An intelligent PID thermal control algorithm dynamically adjusts cooling based on workload, reducing cooling energy consumption by 15–20% compared with conventional fixed-speed strategies.

For maximum system reliability, flexible 3+3 or 5+5 redundant power supplies, real-time power monitoring, and module failure alerts ensure continuous 24/7 operation for mission-critical AI services.

Flexible Expansion for Future Growth

Built with a modular architecture, the G6A80A5 supports up to 12 expansion slots and 16 hot-swappable 2.5-inch NVMe drive bays, enabling high-speed networking, storage acceleration, and scalable heterogeneous AI clusters.

An integrated motherboard and switch board design delivers direct PCIe signal transmission, minimizing latency and signal loss while improving multi-GPU communication efficiency for distributed AI training and inference.

Meet Gooxi at WAIC 2026

As one of the world's leading AI events, WAIC 2026 brings together global innovators to shape the future of artificial intelligence.

Gooxi will showcase the G6A80A5 OAM AI Training & Inference Server at Booth Z2A-203, 2nd Floor, Zhangjiang Science Hall, demonstrating its strengths in open AI accelerator compatibility, unified training and inference, intelligent cooling, and scalable system design.

July 17–20, 2026 | Shanghai, China

Visit Gooxi at WAIC 2026 to discover how next-generation AI infrastructure can accelerate your AI deployment with greater performance, flexibility, and efficiency.