[Edaily Reporter Shin Ha-yeon ] As the scope of artificial intelligence (AI) applications expands from training large-scale models to inference and on-site industrial analysis, data center infrastructure is also becoming more specialized. With the emergence of not only data centers capable of handling large-scale computations but also company-specific dedicated computing environments and the direct deployment of AI infrastructure at the sites where data is generated, competition is intensifying to provide infrastructure tailored to computational scale, data location, and security requirements.

#LG CNS is focusing on building company-specific AI computing environments. Through its container-based AI data center, “AI Box,” the company is establishing physical infrastructure while also introducing “XPUWorks,” a service that allows customers to utilize AI accelerators—such as graphics processing units (GPUs) and neural processing units (NPUs)—as a service.

XPUWorks integrates GPU servers, storage, networking, security, and operational tools to provide an AI development, training, and inference environment tailored to each enterprise’s data and security requirements. A key feature is its ability to configure the necessary infrastructure based on the purpose of AI utilization or computational demand.

KT Cloud provides the infrastructure required for AI workloads as a commercial service. It has commercialized “Colo.AI,” which builds a high-density GPU environment at the Gasan AI Data Center and, based on this, provides AI computing infrastructure such as GPU servers and networks. This service allows AI companies and financial institutions to utilize the necessary computing resources without having to build their own large-scale GPU infrastructure.

iA, Inc.(038880)KT is pursuing a model that involves directly deploying infrastructure to industrial sites requiring AI computing. It has integrated servers, storage, GPUs, NPUs, power and cooling equipment, and operating software into the “NeoCube,” a small, modular AI data center.

The company is currently validating AI computing performance, power and cooling efficiency, and operational stability through a 25-kilowatt (kW) proof-of-concept (POC) and plans to scale up to a 100-kW class in the future. By incorporating domestically developed NPUs from FuriosaAI and Rebellion, the company is targeting the private AI market—including vision and physical AI in manufacturing environments, as well as defense, security, public sector, finance, and healthcare—where external data transmission is restricted or real-time inference is required.

iA, Inc.’s stock price has more than doubled over the past six months, rising from the 1,500 won range to the 3,000 won range. Amid expectations for the expansion of its AI infrastructure business, the stock price rose to 3,390 won on the 10th but has since been experiencing repeated sharp fluctuations.

Meanwhile, as AI applications continue to grow, analysts note that the criteria for competition among data centers are broadening beyond simple computational scale to include data location, security, and operational methods.