Elice Group Unveils K-PMDC-Based AI Full-Stack Strategy

Accelerating the Localization and Ecosystem Build-out of Next-Generation Korean AI Data Center Infrastructure

Elice introduces large-scale clustering and Korea’s first GPU Spot pricing as a CSP, maximizing the cost efficiency of high-performance GPU resources

Jaewon Kim, CEO of Elice Group, announces the company’s 'AI Full-Stack Strategy' at a press briefing.

Elice Group Co., Ltd. (CEO Jaewon Kim), an AI full-stack company providing AI infrastructure, cloud services, and industry-specific solutions, held a press conference titled 'Elice IMPACT 2026: Expanding the Possibilities of the AI Full Stack' on Wednesday, the 15th, at its headquarters in Gangnam-gu, Seoul. During the event, the company unveiled its 'AI Full-Stack Strategy,' which spans the entire spectrum from AI infrastructure to AI education. Elice Group plans to target the next-generation AI cloud infrastructure market with “K-PMDC,” a portable modular data center fully built with domestic technologies, at its core.

In the AI industry, the competitive focus has recently shifted from AI models to computing infrastructure. Securing high-performance GPUs, along with the technological capability to utilize them efficiently, has emerged as a key source of competitive advantage.

Starting from its AI-based education and training platform, Elice Group has expanded into AI cloud infrastructure and evolved into a full-stack player that can directly build and operate across the entire AI value chain. In particular, the company first built AI-dedicated data centers in Korea using an innovative AI PMDC (Portable Modular Data Center) model that previously did not exist in the domestic market. It then developed its own AI cloud virtualization solution, ECI (Elice Cloud Infrastructure), establishing a structure that enables efficient operation and utilization of large-scale GPU resources. On the back of this technological foundation, Elice Group has also proven its execution capabilities by directly building and operating real-world AI infrastructure.

At the briefing, CEO Jaewon Kim stated, “AI infrastructure competitiveness is not about how many GPUs you have, but how well you utilize them,” adding, “Elice Group started as a company specializing in artificial intelligence and software development, and we have independently secured the capabilities to apply infrastructure technologies directly to real services.”

Localizing AI Infrastructure and Building Next-Generation AI PMDC

Elice Group has positioned the localization of AI infrastructure as a core strategy and is moving ahead in earnest with the development of next-generation AI PMDCs. By building AI data centers that reflect next-generation GPU architectures, the company aims to implement infrastructure optimized for high-performance computing environments and establish a foundation capable of responding to rapidly growing AI demand.

In particular, the company has completed the development of a PMDC that supports the next-generation GPU 'Vera Rubin NVL72,' which requires approximately 230 kW of power per rack, thereby acquiring the design capabilities needed to operate stably even in high-power environments.

Elice Group’s AI modular data centers can be built in around three months, compared to the two years or more typically required for conventional data centers. This short deployment timeline and flexible scalability are key strengths. Recently, by partnering with Arista Networks, Elice has gone beyond the traditional, cost-intensive InfiniBand-centric architecture to implement large-scale Ethernet-based clustering technology, securing both cost efficiency and scalability.

The company is also working to build a domestic AI infrastructure ecosystem, not only with global players but also with a wide range of leading Korean companies and institutions across security, energy, networking, and data.

In the security domain, Elice is collaborating with AhnLab to advance next-generation data center network security architectures. In the energy domain, it is working with Standard Energy to explore cooperation in power stability and energy efficiency, while also planning to strengthen collaboration with Lazines (Resinjust/Resingers – confirm correct English brand name as needed) in optical communication solutions for high-performance data centers.

In addition, Elice is partnering with the Korea Institute of Science and Technology Information (KISTI) on the National Data Exchange Node (NDeX) and AI networks; with MakinaRocks on AI transformation in defense and industrial sectors based on ECI; and with LG Uplus on enhancing capabilities in operating large-scale networks and data centers.

Korea’s First GPU Spot Pricing Launched by a Domestic CSP

To maximize AI infrastructure utilization, Elice Group has introduced Korea’s first “GPU Spot” pricing model as a domestic cloud service provider (CSP). GPU Spot pricing leverages idle GPU resources and offers them at up to 50% lower cost compared to on-demand rates. This allows enterprise customers to use the same GPU resources much more cost-efficiently. The plan supports the full lineup of major GPUs, including B200, H100, and A100.

With this new pricing model, Elice Group has become the only CSP in Korea to offer all three pricing options—long-term reserved, on-demand, and Spot—thereby securing a clearly differentiated competitive edge.

Expanding Business Across the Entire AX Spectrum, from AI Infrastructure to Solutions and Education

Under the theme of 'Expanding the Possibilities of the AI Full Stack,' Elice Group is accelerating AX (AI Transformation) not only in AI cloud infrastructure, but also in AI solutions and education.

Elice’s AI document analysis solution lineup, 'Helpy Vision' learns complex and diverse document structures unique to each enterprise and provides technology that converts unstructured data into structured formats. This builds a data foundation that AI agents can actively utilize. In particular, Helpy Vision maintains top-tier performance in Table Structure Recognition (TSR) for complex table layouts.

The company also plans to integrate Helpy Vision into 'AI Helpy Chat,' its generative AI solution tailored for enterprises and educational institutions, to further enhance document comprehension and enable the generation of even more accurate reports. In addition, Elice will integrate field-ready, AI-based UX writing tools and other enterprise-specific utilities, while eliminating the risk of personal and confidential information leakage by running AI Helpy Chat on Elice’s own ML-API-powered LLM.

Building on its track record of providing AX education to major corporations and educational institutions, Elice Group will further advance AX training for enterprise leaders and practitioners. The curriculum will span the full spectrum from generative AI to agent workflow design, with the goal of driving real-world process innovation and job transformation in the AI era. Elice will also offer end-to-end, full-stack support for corporate AI transformation—from company-wide training and AX consulting to PoC implementation and the execution of large-scale projects.

CEO Jaewon Kim concluded, “Elice Group is the only AI full-stack company in Korea that has successfully expanded from an AI-based education and training platform to the construction of AI cloud infrastructure, creating unprecedented efficiency by combining software technologies optimized for GPU hardware,” adding, “By delivering next-generation AI infrastructure built with domestic technologies, we will enhance the self-sufficiency of Korea’s AI ecosystem and lay the foundation for companies to achieve AI innovation without being constrained by cost barriers.”