LG, Nvidia Deepen AI Partnership with New Co-Development Alliance

LG AI Research and NVDA Nvidia formalized a co-development alliance on April 22, 2026, agreeing to jointly develop next-generation AI models by combining LG's Exaone and Nvidia's Nemotron model families, targeting specialized applications in manufacturing and biotech. The agreement was signed in Seoul and expands a collaboration that traces back to LG's Exaone 3.0 development cycle, now extending to the latest multimodal Exaone 4.5 release.

LG AI Research is leveraging Nvidia's full software stack — GPUs, the NeMo development framework, and TensorRT-LLM inference software — to optimize training and deployment of its domain-specific models. LG plans to use Nvidia's high-quality training data to enhance Exaone's performance in verticals where general-purpose models underperform, particularly in Korean-language industrial applications and life sciences workflows.

The deal reflects the global trend toward sovereign AI, where nations invest in domestic model capabilities rather than fully depending on US-developed systems. South Korea ranked third globally in notable AI models in Stanford's 2026 AI Index, with four of its five selected models from LG's Exaone lineup. For Nvidia, locking in Exaone's training and inference infrastructure on its hardware insulates Korean enterprise revenue against potential incursion from GOOGL Google TPUs — whose South Korean partnerships are also expanding — and other alternative silicon providers.