 |
- Trillion Labs’ industrial world models capture how AI Factories — and the plants that power them — actually behave and decide how they should run: the intelligence layer that complements NVIDIA Omniverse’s physics-accurate digital twins.
SEOUL, South Korea, June 8, 2026 /PRNewswire/ — Trillion Labs, a South Korean foundation model lab, announced that it is developing Industrial World Models for AI Factories, built with NVIDIA Omniverse libraries and NVIDIA Nemotron open models.
Industrial World Models are designed to enable AI systems and agents to understand, simulate, and optimize complex industrial environments such as AI data centers and power plants. Trillion Labs builds its world models from scratch, using NVIDIA Omniverse libraries for physics-accurate simulation and NVIDIA Nemotron for advanced reasoning, to serve as the intelligence layer for next-generation AI Factories.
Trillion Labs has built proprietary foundation models spanning large language models (LLMs) and vision-language models (VLMs), and is now developing world models for physical infrastructure. The company has also developed expertise across the entire AI stack, from training and inference to deployment and operations. This full-stack approach allows Trillion Labs to design and optimize models specifically for real-world industrial applications.
While continuing its foundation model research, Trillion Labs is increasingly focusing on industrial and engineering-specific AI models. Rather than developing general-purpose or generic AI, the company aims to build specialized foundation models capable of understanding and manipulating complex physical infrastructure and industrial operations.
Trillion Labs’ key focus is AI Factories, where data centers, power generation facilities, cooling systems, and other operational infrastructure operate in tightly connected ecosystems — increasingly, under one roof. Trillion Labs is developing Industrial World Models that can understand complex schematics, time-series data, facility operations, maintenance histories, operational constraints, and infrastructure dependencies, enabling AI systems to support optimization across industrial environments.
The company has been collaborating with GS Group on critical energy infrastructure, as well as other enterprise partners across industrial domains.
“Large language models remain the foundation of advanced intelligence, but understanding language alone is not enough to understand the physical world,” said Jay Shin, CEO of Trillion Labs. “Industrial World Models are our next step toward extending AI from language understanding to real-world reasoning. By combining foundation models we build ourselves with simulation and specialized reasoning, we aim to build AI systems capable of understanding and optimizing critical infrastructure at scale.”
The initiative reflects Trillion Labs’ broader vision of Industrial Superintelligence, where AI systems can understand, simulate, help, and eventually autonomously operate complex industrial systems. The company believes Industrial World Models will become a foundational building block for the next generation of AI-powered infrastructure.
Shin presented on Korean LLMs as a speaker at NVIDIA GTC 2025 and has maintained a meaningful partnership with NVIDIA through the Inception program. Trillion Labs views Industrial World Models as a key opportunity to combine its foundation model expertise with NVIDIA’s AI factory, digital twin, and physical AI ecosystem.
As the build-out of AI Factories continues to explode globally, Trillion Labs aims to establish a new category of Industrial Intelligence that bridges foundation models and real-world infrastructure.
About Trillion Labs
Trillion Labs is an AI research lab founded in 2024 by Jay Shin, a key architect behind NAVER’s HyperCLOVA X. Within a year of its founding, the company developed and open-sourced a full lineup of large language models (LLMs)—including 7B, 21B, and 70B parameter models—from scratch.
Since then, Trillion Labs has expanded beyond language models, introducing a vision-language model (VLM), Korea’s first large language model built on a diffusion-based transformer architecture (dLLM), and gWorld, a mobile world model. Through these advancements, the company has demonstrated full-stack AI capabilities spanning foundation models, multimodal AI, and world models.