South Korean Telecom LG U+ to Launch Enterprise AI Token Business
South Korean telecom operator LG U+ announced plans to launch an AI token business next year to help enterprises reduce AI transformation costs using proprietary optimization techniques. The service combines harness engineering for agent tasks, intelligent routing, and heterogeneous semiconductor combinations. This move marks the entry of all three major South Korean telecom operators into the token economy, intensifying competition in the B2B AI infrastructure space. By leveraging cost-reduction strategies like heterogeneous hardware and model collaboration, it addresses the high expenses traditionally associated with enterprise AI adoption. LG U+ utilizes a chip combination strategy pairing NVIDIA and AMD GPUs with FuriosaAI NPUs to separately handle problem analysis and answer generation tasks. Additionally, a small-large model collaboration method achieves a token generation speed more than twice as fast by having smaller models quickly generate answers for larger models to verify.
## BACKGROUND
AI Token businesses focus on providing optimized language model access, infrastructure, and API pricing models for enterprises transitioning into AI-driven operations. Harness engineering involves building controlled environments and task execution pathways for AI agents, while heterogeneous hardware combines different types of processors—such as GPUs and NPUs—to maximize efficiency for specific computational workloads.