AI full-stack company Elice Group held a media briefing at its headquarters in Seoul’s Gangnam district on Tuesday and announced a business strategy covering areas from AI infrastructure to education.

The core of the strategy is a domestically developed "portable modular data centre (PMDC)." It can cut the time needed to build a data centre from more than 2 years to about 3 months. The company also completed development of a PMDC that supports Nvidia’s next-generation GPU, Vera Rubin NVL72, which requires 230 kW of power per rack. Recently, it partnered with Arista Networks to implement Ethernet-based large-scale clustering technology, moving beyond a high-cost structure centred on InfiniBand, and securing cost efficiency and scalability at the same time.

Elice Group also introduced a "GPU spot pricing plan," the first such offer by a domestic cloud service provider (CSP). It uses idle GPU resources to provide services at costs of up to 50 percent of on-demand pricing, and supports a full lineup of major GPUs including B200, H100 and A100. The company said it is the only domestic CSP with all three billing models: long-term contracts, on-demand and spot.

It is also expanding cooperation to build a domestic AI infrastructure ecosystem. It will pursue cooperation with AhnLab to advance next-generation data centre network security, with Standard Energy on power stability and energy efficiency, and with Resengers on optical communications solutions for high-performance data centres. With the Korea Institute of Science and Technology Information (KISTI), it will work on the National Data Exchange Node (NDeX) and AI networks. With MakinaRocks, it will pursue defence and industrial AI transformation based on Elice Cloud Infra (ECI). With LG Uplus, it will cooperate to strengthen capabilities in operating large-scale networks and data centres.

It is also expanding its presence in solutions and education. Through its AI document analysis solution, the "Helpie Vision" lineup, it will provide technology to structure unstructured data and plans to integrate it with its generative AI solution "AI Helpie Chat" to improve document comprehension and the accuracy of report generation. It will also expand enterprise AX training across areas from generative AI to building agent workflows.

Elice Group CEO Jaewon Kim (김재원) said AI infrastructure competitiveness depends not on the number of GPUs but on how well they are used. He said the company will raise the self-reliance of the domestic AI ecosystem with next-generation AI infrastructure built with domestic technology and create a foundation for companies to achieve AI innovation without cost barriers.