Cango Launches Commercial AI Inference Operations at 50-MW Georgia Facility

The new subsidiary, EcoHash, uses standardized "plug-and-play" modules integrated with a proprietary software layer called the EcoLink Orchestration Platform. This system is designed to aggregate geographically distributed computing capacity and provide failover capabilities for artificial intelligence (AI) developers and data center operators. Cango (NYSE: CANG) stated that the Georgia site will serve as a showroom to demonstrate performance across various thermal and power configurations while providing commercial services.

The move comes as Cango seeks to diversify its energy-intensive infrastructure. The company cited research from Goldman Sachs indicating that U.S. data center power demand could reach 700 TWh by 2030, creating a potential 400 TWh gap between demand and available infrastructure. By repurposing existing mining sites, Cango aims to provide low-latency compute power closer to data sources for AI inference workloads.

Jack Jin, Chief Technology Officer of EcoHash, said the orchestration layer is intended to connect decentralized energy assets with the requirements of large language model (LLM) inference and generative AI. Cango plans to invite global partners to integrate with the EcoHash network and intends to replicate the Georgia operational model at other locations within and outside its current network.