LG Uplus plans enterprise AI optimization business for 2027
The AMW Read
The planned enterprise offering adds a telecom operator to the inference optimization market, but its reported savings remain workload-specific company claims.
LG Uplus plans enterprise AI optimization business for 2027
LG Uplus plans to expand into business-to-business AI token optimization in 2027, using partnerships and its own infrastructure. The company says an internal demonstration cut coding-agent operating costs by about 90% compared with a conventional model. It also reported that model routing reduced general-query costs by more than 25% and coding-agent costs by more than 30%. CTO Lee Sang-yeop cautioned that savings vary substantially by task.
The plan places a telecom operator in the AI infrastructure market, where providers compete to reduce the cost of completing a task rather than only the price of each token. LG Uplus says it will combine smaller models, inference optimization and different chips. It also reports that a smaller on-device model retained more than 95% of the performance of a cloud model used for comparison. The reported results are company claims, and the source does not provide enough detail to assess their general applicability.
For enterprise buyers and builders, the practical test will be whether savings hold for their own workloads while maintaining the required output quality. LG Uplus says its routing system selects models based on query difficulty and task type. It is also working with OTAI on Nvidia technology and MangoBoost on AMD technology, while evaluating FuriosaAI processors. The planned business enters a market where KT has launched a token-management offering and SK Telecom is pursuing a token gateway.