Compact Yet Powerful AI Computing Center... Baro AI Says "Deployable in Any Urban Building or Research Facility"

Baro AI Opens 'Baro Space' in Pyeongtaek, Gyeonggi Province
Standardized Cell Units (Up to 400 GPUs) Enable Scalability
Preparing GPUaaS Business and Pursuing Global Expansion

Baro AI, a company specializing in artificial intelligence (AI) infrastructure, has unveiled a next-generation AI data center based on liquid cooling that can be operated on a small scale within urban buildings or research facilities.

Baro AI's AI Computing Center 'Baro Space' opened on the 29th in Pyeongtaek, Gyeonggi Province. Provided by Baro AI

On October 29, Baro AI officially opened its AI computing center, 'Baro Space,' in Pyeongtaek, Gyeonggi Province. Baro Space features a standardized GPU server system using liquid cooling, organized into cell units (up to 400 GPUs per cell), allowing for flexible scalability.

Unlike hyperscale data centers, which are restricted by location and require significant time and cost to establish, Baro Space is expected to be quickly installed and operated in AI research environments or urban settings of any kind.

Baro Space is an implementation of Baro AI's independently developed HACC (Hyper Modular AI Computing Center) architecture, representing a compact yet high-performance cell-based AI computing center.

It can provide computing resources directly to sites with AI demand, such as companies, universities, research institutes, and knowledge industry centers. Additionally, its cell units can be replicated and expanded like Lego blocks as needed, maximizing efficiency. As long as 250 to 500 kW of power is secured, it can be installed within existing buildings, ensuring stable operation even in urban buildings or research facilities.

Baro Space is fully equipped for a GPU-as-a-Service (GPUaaS) environment, and the company has developed its own integrated management system to monitor and optimize real-time power usage, temperature, and cluster status.

According to Baro AI, its self-developed server 'Poseidon' maintains a low noise level of 39dB even under full load, and can perform long-term training at GPU temperatures of 50 to 60 degrees Celsius without any performance degradation. This is thanks to its patented liquid cooling technology, which reduces power consumption by 30 to 35 percent compared to air-cooled systems and provides eco-friendly benefits by lowering carbon emissions.

Poseidon has been supplied to major universities, hospitals, and research institutes for use in various projects. Notably, a research team from Konkuk University College of Medicine, using Baro AI's infrastructure, won first place at the IEEE (Institute of Electrical and Electronics Engineers) Global AI-Based Alzheimer’s Assessment Competition, demonstrating the company's technological prowess.

Baro AI plans to roll out an export-oriented HACC model that can be rapidly deployed in regions with underdeveloped energy infrastructure, such as Southeast Asia, the Middle East, and Latin America. The company is promoting AI infrastructure supply projects linked to government aid funding, aiming to establish a global sovereign AI network through the export of Korea's AI technology and infrastructure.

Lee Yongdeok, CEO of Baro AI, stated, "The world is focusing on hyperscale data centers, but considering the rapid pace of GPU generation upgrades and utilization risks, centralizing all AI workloads is inefficient." He added, "Starting small and expanding according to demand, while operating with a high-efficiency liquid cooling structure to reduce operating costs, is a practical alternative."

This content was produced with the assistance of AI translation services.