AI Models & Platforms

Cango’s EcoHash Begins Commercial GPU Compute at Georgia AI Facility

mm
Add Unite.AI to your preferred sources on Google

Cango Inc. said its high-performance computing and AI inference subsidiary, EcoHash Technology LLC, completed dedicated AI infrastructure modifications at its Georgia facility and began commercial GPU compute services with an initial batch of GPU servers, the company announced on September 3, 2026.

EcoHash completed the infrastructure modifications and retrofits for the dedicated AI compute section within Cango’s owned 50MW Georgia facility. According to the company, that section supports up to 3 megawatts of capacity, with scope for future expansion.

Container Deployment and Site Design

Part of the initial deployment involved the delivery, installation, and testing of high-density compute containers, which began their operational rollout through the power-on and activation of an initial batch of GPU servers. The containerized data center solution is designed as a modular building block for AI compute, and the company said that by standardizing power, cooling, networking, and operations into a repeatable unit, it can scale AI infrastructure faster and more cost-effectively across suitable sites.

The site is built around a series of standardized, high-density AI compute modules, each engineered to support sustained high-density GPU loads. Power is delivered through an industrial-grade transformer setup, a precision cooling system supports the compute density, and fiber infrastructure provides network connectivity for AI compute workloads.

Because the compute modules were deployed alongside Cango’s existing mining infrastructure at the same Georgia site, EcoHash was able to minimize new construction and site retrofit work, which the company said accelerated the path from module delivery to power-on.

Commercial Services and Customers

On the commercial side, EcoHash has begun providing commercial GPU compute services using part of its initial batch of GPU servers now in operation, with additional GPU deployment ongoing. The company said this marks the initial commercial utilization of EcoHash’s GPU capacity at the Georgia site. Initial and prospective customers include GPU cloud platforms and AI-native cloud infrastructure providers, and Cango said it is in discussions with additional prospective clients across other categories, including decentralized compute networks and advanced AI research laboratories.

“Powering on our first batch of GPU servers and beginning commercial GPU compute services marks an important step for EcoHash,” said Simon Tang, chief financial officer of Cango. He described the Georgia rollout as a strategic proof-of-concept hub that the company intends to replicate at future sites.

Prior Buildout and Business Models

The September 3 milestones follow buildout progress Cango detailed in its second-quarter 2026 results, released August 31, 2026. In that report, chief executive officer Paul Yu said the Georgia site completed conversion in early July 2026, that container units had been delivered and installed on site, and that GPU hardware had been procured and was arriving in staged batches to support a phased ramp-up.

Yu also outlined two intended business models for the AI compute operation. The first is bare-metal GPU hosting, using the company’s infrastructure to offer a standardized deployment environment. The second is colocation, intended to enhance overall infrastructure utilization. He said the Georgia site was in the process of onboarding customers, with revenue expected to be recognized in the third quarter, and noted that Cango had begun operating test nodes in Texas and on the West Coast to support customers requiring proximity-based deployment.

Cango’s second-quarter revenue totaled US$50.8 million, driven primarily by US$47.4 million from its Bitcoin mining business, with a net loss of US$81.6 million that the company attributed mainly to non-cash impairment and disposal losses on mining machines.

EcoHash Background

EcoHash Technology LLC was founded in 2025 as Cango’s dedicated high-performance computing and AI inference arm. Cango first announced the subsidiary’s commercial operations on April 13, 2026, alongside the launch of its digital portal, and said at the time that it would dedicate space at the Georgia facility as a demonstration site for its plug-and-play compute modules.

Cango is a Bitcoin mining company whose operations span North America, the Middle East, South America, and East Africa. Since entering the digital asset space in November 2024, the company has activated pilot projects in integrated energy solutions and distributed AI computing, and continues to operate an online international used car export business.

Theo Nash is an AI-generated specialist at Unite.AI, covering AI infrastructure, compute, and the hardware systems that power modern artificial intelligence. His work focuses on the technical foundations behind large-scale AI workloads, including data centers, accelerators, networking, and the software stacks that tie them together.

With an analytical and engineering-driven perspective, Theo examines how advances in GPUs, custom silicon, memory architectures, and distributed systems enable new generations of AI models. He pays particular attention to performance trade-offs, energy efficiency, scalability, and the practical constraints that shape real-world deployment of AI infrastructure.

Articles authored by Theo Nash are AI-generated and reviewed by Unite.AI’s editorial team to ensure technical accuracy, clarity, and responsible coverage of the rapidly evolving AI compute landscape.