LG AI Research EXAONE: Expanding Group-Scenario Models to External Markets
1) Licensing of EXAONE foundation models and industry-specific versions; 2) Project-based or subscription fees for scien
Key Fields
FIELD STAMPS📌 Background
Established in 2020, LG AI Research leverages real-world industrial scenarios from LG Group—including chemicals, biotechnology, and electronics—to train its proprietary EXAONE model. In September 2026, during the AI Talk Concert, LG announced the expansion of EXAONE's external supply. Its K-ExaOne model ranked in the top ten in global performance evaluations, making it the only Korean model on the list, with enterprise-grade AI applications becoming its commercial focus.
👤 Target Customers
Large and medium-sized enterprise clients in sectors like chemicals, materials, pharmaceuticals, and finance that require vertical AI capabilities, as well as Korean companies needing specialized Korean-language models.
💰 Revenue Streams
1) Licensing of EXAONE foundation models and industry-specific versions; 2) Project-based or subscription fees for scientific discovery platforms like EXAONE Discovery; 3) Joint R&D of new materials and ingredients with external partners, sharing commercialization profits.
🧮 Cost Structure
High costs for large model training compute and data; R&D and scenario validation costs shared within the group; investment in research talent and academic partnerships.
🛡️ Moat
A unique training and validation loop built on real data and application scenarios from LG Group's chemical, biological, and electronic industries; scarcity of specialized Korean-language domain models; brand and compliance endorsement through partnerships with organizations like UNESCO.
🔑 Keys to Success
- Transforming internal group validation results into standardized, sellable products
- Establishing quantifiable evidence of cost reduction and efficiency gains in vertical scenarios like materials and biomedicine
- Maintaining foundation model performance at the global forefront
⚠️ Risks
- Internal group scenario experience may not necessarily adapt to external client industries
- Mismatch between high R&D investment and external revenue
🏢 Cases
- Collaborated with LG Household & Health Care to screen the anti-hair loss ingredient Rhamsydil from over 420,000 candidate compounds in one day
- Jointly developed immersion cooling fluid materials for data centers with GS Caltex
- Announced the expansion of EXAONE's external supply at the LG AI Talk Concert in September 2026, entering the enterprise AI market
📊 SWOT Analysis
Strengths
- Unique validation data provided by internal group scenarios in chemicals, biology, and manufacturing
- K-ExaOne ranked in the global top ten, the only Korean model to do so
- Demonstrable success cases, such as screening anti-hair loss ingredients from 420,000 compounds in a single day
Weaknesses
- Global influence and compute investment lag behind US giants like OpenAI and Google
- Commercialization started relatively late, with a limited number of external client cases
Opportunities
- Growing demand for enterprise-grade vertical AI, with high willingness to pay in material and drug discovery scenarios
- Preference among Korean and Asian enterprises for localized, controllable models
Threats
- Performance gains in open-source models compressing the space for licensing
- Intensifying competition in the Korean enterprise market from US and Chinese large model providers