K-Deepseek Success: AI Infrastructure Software Capabilities Key

AI Global Conference: MOREH Draws the Future of Korean-style Deepseek Amidst the fervor of the 'AI Global Conference' hosted by the Ministry of Science and ICT, AI infrastructure solution company MOREH captured the industry's attention by presenting an innovative blueprint for realizing 'Korean-style Deepseek'. The event, held on the 25th at Lotte Hotel in Jung-gu, Seoul, brought together global AI leaders such as NVIDIA, AMD, and OpenAI, becoming an important forum to discuss the future of AI technology. Under the theme 'Computing Infrastructure and AI Models: Seize the Initiative in Innovation,' the latest AI trends and development prospects were shared, followed by keynote speeches from overseas big tech companies and presentations on government plans to strengthen AI capabilities. In particular, the AI semiconductor, AI data center, and AI model and service sessions featured a comprehensive lineup of major domestic AI companies, showcasing a diverse technological feast. In the AI Model and Service session, Cho Hyung-geun, CSO (Chief Strategy Officer) of MOREH, who gave the presentation, identified 'AI Infrastructure Software (SW)' as the key to the success of 'Deepseek,' which has opened new horizons in AI model development. He emphasized that the process of bundling thousands of GPUs into a single system to realize hyperscale AI is entirely within the domain of software, and that the role of AI infrastructure SW is fundamentally changing. CSO Cho asserted that MOREH, which has been dedicated to developing AI infrastructure SW that surpasses NVIDIA CUDA since its inception, holds the optimal solution for Korean-style Deepseek. MOREH's solution offers perfect compatibility with heterogeneous GPU and NPU resources from companies like NVIDIA and AMD, along with unique GPU virtualization technology, providing AI businesses not only cost and time efficiency but also a wide range of options, thereby opening up new possibilities. MOREH's technological prowess is also recognized internationally. Recently, GCV, a UK-based venture-focused media, listed MOREH among the 'Global AI startups to watch with the emergence of Chinese Deepseek,' evaluating it as the company closest to the Deepseek approach through its proprietary software technology. Established in September 2020 by the key developers of Seoul National University's 'Cheondung' supercomputer, MOREH is dramatically lowering the entry barrier for AI model development by automatically implementing the parallelization and optimization techniques required for large-scale AI model development with its fully automated compiler technology. In practice, MOREH has demonstrated its capabilities by building an AMD Instinct GPU-based AI data center to support KT AI cloud services. Last December, it further solidified its technological leadership by open-sourcing its self-developed Korean LLM foundation model, 'Motif,' on Hugging Face.