Cango's subsidiary EcoHash has launched a commercial AI inference platform, targeting developers and compute operators seeking low-latency, near-source capacity. The move comes as Goldman Sachs researchers project U.S. data center power demand could hit 700 TWh by 2030, driven largely by AI workloads—far outpacing current supply of 300 TWh.

The company is repurposing part of its 50MW Georgia mining facility as a proof-of-concept hub for AI compute infrastructure. EcoHash aims to bridge what CEO Paul Yu calls the 'Power Gap'—the widening disconnect between AI's energy needs and grid capacity constraints.

Chief Technology Officer Jack Jin positioned EcoHash as Cango's vehicle to build distributed, future-ready infrastructure. The platform offers modular solutions for energy-intensive operators underserved by traditional data centers.