South Korea’s LG CNS will invest ₩381.4 billion (approximately $275.3 million) to secure GPUs and infrastructure equipment, including Nvidia’s next-generation AI accelerator Vera Rubin. The move goes beyond simple equipment procurement—it signals a strategic shift toward preemptively securing group-level AI factory construction capabilities and transitioning to a business model that combines GPU subscription services with data center design-build-operate (DBO) offerings.

According to filings with South Korea’s Financial Supervisory Service electronic disclosure system on the 17th, LG CNS held a board meeting on the 11th and approved the acquisition of GPUs and infrastructure equipment, including Nvidia Vera Rubin units, for ₩381.4 billion (approximately $275.3 million). The counterparty is Hewlett Packard Korea Ltd., among others, with the acquisition scheduled for June 30 next year. The investment equals 7.2% of the company’s consolidated total assets of ₩5.2853 trillion (approximately $3.8 billion) as of end-2025.

The company stated the purpose of the investment is to secure GPU equipment for AI factory construction. It follows the memorandum of understanding signed between LG Group and Nvidia in August this year, aimed at supporting AI factory development and physical AI advancement.

The scale of the investment stands out even more when compared with historical spending. According to KB Securities analysis, ₩381.4 billion is approximately 8.6 times LG CNS’s typical annual tangible asset acquisitions. This reinforces the interpretation that the move is primarily a preemptive investment for future large-scale AI factory projects rather than simple GPU procurement.

LG Group plans to build a Vera Rubin-based reference site in the first half of next year to validate integrated cooling, power, and IT technologies, followed by an 80-megawatt (MW) AI factory in Cheonan, South Chungcheong Province, by the first half of 2028. KB Securities expects LG CNS to secure a standardized architecture covering the entire AI factory lifecycle—from design and virtual commissioning to construction and operations.

A Vera Rubin-based AI factory integrates not only GPUs but also networking, storage, power, liquid cooling, digital twin technology, and operational software into a single infrastructure. Unlike conventional data centers, it is a composite infrastructure specialized for AI computing, making design and operational expertise highly valuable.

The acquired infrastructure will also directly connect to LG CNS’s AI infrastructure services business. Since launching its proprietary platform XPUWorks on July 15, LG CNS has been offering enterprises various AI computing resources on a subscription basis, allowing customers to access computing power as a service without purchasing expensive GPUs outright.

An LG CNS representative explained that XPUWorks equipment “provides computing resources equipped with Nvidia B300 and H200, and the GPUs are LG CNS assets,” adding that “neural processing units (NPUs) and data processing units (DPUs) are currently under validation.” The service currently targets electronics and manufacturing companies, with plans to expand to financial customers.

The development trajectory of XPUWorks is also visible in regulatory filings. The “GPU-as-a-Service (GPUaaS) technology research” item first appeared in the 2025 annual report, evolved into “GPUaaS technology research and platform development” in the first-quarter report this year, and the XPUWorks name was added in the semi-annual report.

Analysts see potential for the scope to expand beyond in-house GPU services to customer-dedicated private GPU-as-a-Service (GPUaaS) offerings and AI factory DBO projects. The company could broaden its business model from project-based data center construction to a recurring revenue structure that combines post-construction operations and services.

An LG CNS representative said, “We currently have projects linking XPUWorks to DBO customers,” adding, “We expect these linkages to expand further as we scale the business at the AI factory level.”

AI demand within LG Group is also expected to support initial GPU utilization rates. The model involves securing baseline utilization through affiliate AI computing demand—spanning robotics, mobility, and AI model development—then supplying remaining GPU capacity to external customers. Group demand effectively serves as an anchor tenant, reducing the risk of external business expansion.

Kim Jun-seop, an analyst at KB Securities, said, “We view LG CNS’s investment as a preemptive reference investment for future large-scale AI factory projects rather than simple GPU procurement,” adding, “If the validated construction and operations model and platform can be applied repeatedly, there is potential to expand the business from project-based revenue to recurring revenue combining operations and services.”

Competition in AI infrastructure investment among South Korean IT services firms is also accelerating. Samsung SDS is building an AI data center in Gumi to accommodate GPU subscription services and AI business expansion, with plans to invest a total of ₩463.9 billion (approximately $334.9 million) from 2024 to 2029. Naver invested ₩741.1 billion (approximately $535.0 million) in servers and equipment in the first half of this year, an increase of ₩404.7 billion (approximately $292.2 million) compared with the same period last year.

However, the three companies differ in investment targets and timelines. LG CNS disclosed the amount for GPUs and infrastructure equipment to be acquired by June next year, while Naver reported first-half server and equipment investment results including non-GPU hardware. Samsung SDS’s investment figure covers data center construction costs through 2029.

LG CNS’s latest investment is being viewed as a signal of its transformation from a simple equipment buyer to an integrated player spanning design, construction, operations, and services in the AI infrastructure market. By securing the latest Vera Rubin GPUs, the company is positioning itself to establish a first-mover reference for AI factory construction, with a two-track strategy targeting both group internal demand and external customers now taking visible shape.