🧠 Food for thought

Implications, context, and why it matters.

LG is building out AI infrastructure alongside its models

• EXAONE 4.5 sits within a wider AI plan that also funds large-scale infrastructure 1.
• The “One LG” initiative includes the Paju AI data center, a 200-megawatt site built for up to 120,000 graphics processing units (GPUs), with completion set for 2027 1.
• LG treats the site as the hub for “One LG,” linking LG Electronics, LG Energy Solution, and LG CNS, LG’s IT services arm 2.
• LG AI Research also built K-EXAONE, an open-weight large-scale multilingual language model. It uses a Mixture-of-Experts architecture, which turns on only part of the model per task. K-EXAONE has 236 billion total parameters with 23 billion active during inference 3.

LG’s AI strategy emphasizes industrial deployment and agentic systems

• LG sets the goal as AI that supports human life and addresses real-world challenges 2. It rejects a focus on “the height of intelligence” 2.
• LG is building “agentic AI” that cycles through planning, execution, evaluation, then revision 2. It goes beyond one-off query responses 2.
• LG expects EXAONE 4.5 vision-language model technology to serve as the brain for “KAPEX,” a Korean humanoid robot 2. The company calls this direction “physical AI,” which places AI systems inside machines that interact with the real world 2.
• K-EXAONE can call tools through OpenAI- and Hugging Face-style tool specifications, which helps support agentic use cases 3.