LG and Nvidia Partner to Develop AI Models, Expand EXAONE Ecosystem in South Korea

Gate News message, April 22 — LG Group and Nvidia announced a partnership on April 22 to jointly develop new AI models and expand the EXAONE ecosystem, as part of South Korea's government-backed "Proprietary AI Foundation Model" project.

The collaboration will integrate LG's EXAONE large language model with Nvidia's Nemotron open ecosystem to create domain-specific models. EXAONE 4.5, unveiled by LG in early April, features 33 billion parameters and operates under a non-commercial license, limiting its use to research and education; commercial deployment requires a separate agreement. The partnership also leverages Nvidia's NeMo framework for model development and training.

South Korea's sovereign AI initiative, backed by approximately $358 million in government funding, aims to build globally competitive AI capabilities by 2027. The project includes plans for a 120,000-GPU data center in Paju and national deployments of over 250,000 Nvidia GPUs across sovereign clouds and AI infrastructure, positioning state-backed models to compete through regional language optimization and specialized enterprise applications such as technical document analysis.