 |
Lee Sang-yeop, Chief Technology Officer (CTO) of LG Uplus, and Lim Woo-hyung, co-head researcher at LG AI Research Institute, speak at a press briefing in Barcelona, Spain, on March 1 (local time), a day before the opening of Mobile World Congress (MWC) 2026. (Photo: Lee Joonhyun, Alpha Biz) |
[Alpha Biz = Reporter Lee Joonhyun, Barcelona] LG Uplus and LG AI Research announced plans to build a large-scale Artificial Intelligence Data Center (AIDC) in Paju, Gyeonggi Province, capable of housing up to 120,000 graphics processing units (GPUs), with completion targeted for 2027.
The infrastructure plan was unveiled by Lee Sang-yeop, Chief Technology Officer (CTO) of LG Uplus, and Lim Woo-hyung, co-head researcher at LG AI Research, during a press briefing held in Barcelona, Spain, on March 1 (local time), ahead of Mobile World Congress (MWC) 2026.
The Paju AIDC will have a capacity of 200 megawatts (MW) and will serve as a central base that integrates the technological capabilities of various LG affiliates, including LG Electronics, LG Energy Solution and LG CNS.
LG Uplus plans to develop the facility into a hybrid infrastructure combining enterprise AI platforms with neural processing units (NPUs) developed by FuriosaAI.
Alongside the infrastructure expansion, the company also revealed improvements in AI model inference efficiency.
Lim Woo-hyung said the K-EXAONE model adopted a Mixture-of-Experts (MoE) architecture, activating only about 10 percent of its 236 billion parameters during inference, which improved inference efficiency by an average of 30 percent compared with previous models.
Lee Sang-yeop said the surge in data traffic is a key factor behind the large-scale infrastructure investment.
“Voice-based V2V (Voice-to-Voice) services generate more than 30 times the traffic compared with text-based services, making a cost-efficient hybrid AI infrastructure essential,” Lee said.
He added that the company has successfully tested running a 1.2-billion-parameter small language model (SLM) directly on devices without relying on cloud computing, significantly reducing power consumption.
Lee also noted that LG Uplus is preparing for the commercialization of 5G Standalone (SA) and the adoption of SRv6 routing technology to enable low-latency AI-native networks.
Alphabiz 이준현 기자(wtcloud83@alphabiz.co.kr)
