EcoHash Technology LLC, the dedicated HPC and AI inference subsidiary of Cango Inc. (NYSE: CANG), launched its public digital portal on 13 April 2026, announcing the start of commercial operations.
It also unveiled plans to operate a portion of its 50-megawatt (MW) Georgia mining facility as a live proof-of-concept hub for the AI compute industry.
What is EcoHash, and why is it entering the market now?
Cango (CANG) founded EcoHash in 2025 as part of its goal to convert the company’s global energy infrastructure into a distributed AI compute network. EcoHash’s commercial launch targets AI developers seeking low-latency, near-source compute capacity, and energy-intensive compute operators looking for modular pathways to infrastructure diversification. Cango (CANG) believes the latter is underserved by conventional data center providers.
This development is coming at a time when researchers from Goldman Sachs are forecasting that U.S. data center power demand could reach 700 TWh by 2030, and this will be driven predominantly by AI inference workloads.
However, the current available supply remains just above 300 TWh, leaving a gap of about 400 TWh even as compute demand steadily increases.
This is the commercial rationale EcoHash is built around, and it was pointed out by Cango’s CEO Paul Yu, who calls the “Power Gap” the disconnect between rising AI compute demand and constrained grid capacity.
According to Jack Jin, chief technology officer of EcoHash, “EcoHash represents the core vehicle of our strategy to architect a future-ready platform and serve as our next growth engine, now entering a phase of accelerated commercialization.”
The subsidiary’s commercial launch follows a period of intensive capital deployment. In April 2026, Cango (CANG) announced the completion of two financing transactions totaling $75 million, a $65 million equity close from board insiders Xin Jin and Chang-Wei Chiu, and a $10 million convertible note from Hong Kong-listed DL Holdings Group Limited (HKEX: 1709).
Cango (CANG) also entered a memorandum of understanding with DL Holdings for up to $10 million in further co-investment.
Those transactions followed an earlier $305 million boost from the sale of Bitcoin holdings used to retire debt and reset the balance sheet.
What is the Georgia facility designed to demonstrate?
EcoHash’s launch strategy is backed by the Cango-owned 50MW Georgia mining facility, where the company is dedicating space to operate full-series container models as what it describes as a “living showroom”.
The site is engineered to demonstrate real-world performance across varying thermal and power configurations, functioning as a strategic proof-of-concept hub for industry collaborators across the digital infrastructure and mining ecosystem.
Cango (CANG) intends for a portion of the Georgia facility to serve as the replicable template for a globally distributed AI compute network, with ambitions to scale the model across high-potential sites both within and beyond its existing mining locations spanning North America, the Middle East, South America, and East Africa.
The commercial viability of its plug-and-play modules in Georgia will enable the company to attract global partners into the EcoHash network, operators who can integrate existing infrastructure into the platform instead of building new data centers from scratch.
How does the EcoLink platform come into the picture?
The operational backbone of EcoHash is the proprietary EcoLink Orchestration Platform, a software layer that unifies and schedules geographically dispersed compute capacity across the network.
EcoLink is built to deliver enterprise-grade uptime through intelligent failover, provisioning compute power to meet real-time workload demands.
It is the mechanism that transforms a collection of repurposed mining sites into something resembling a conventional hyperscale offering.
In his comment, Jin stated that EcoLink is “the central nervous system of our network”, built to enable intelligent, real-time resource allocation connecting decentralized energy assets directly to the demands of large language model inference, generative AI, and a growing range of compute-intensive applications.
The result, per Cango (CANG), is elastic, low-latency compute that scales on demand, without the capital expenditure and multi-year lead times associated with building new data center capacity.
