AI news ·
LG Uplus forms alliance with GS E&C and others to develop modular AI data centers
LG Uplus formed an alliance with four firms to standardize prefabricated modular data centers for AI, targeting a 100MW hyperscale design and a retrofit model.
LG Uplus has formed a consortium with four South Korean construction and design firms to develop a standard model for prefabricated modular data centers (PMDCs) aimed at AI workloads. The telecom operator announced the PMDC Alliance on September 6 at its Pyeongchon Megacenter in Anyang, bringing together partners who will handle everything from technical requirements to on-site assembly.
Who's in the alliance
The group includes GS Engineering & Construction, Xi C&A, Gansam Architects, and D&O CM. LG Uplus will set operating standards and technical requirements and verify the final model. The partner firms will optimize construction, design, MEP (mechanical, electrical, and plumbing) work, and project and construction management.
What the group is building
Since April, the alliance has reviewed two models. The first is a standard design for a new 100MW hyperscale data center. The second is a retrofit model for converting existing buildings into data centers. Both approaches rely on standardizing building structures, power systems, and cooling facilities so components can be prefabricated in factories and assembled on site.
This method shortens construction timelines compared to traditional builds. It also allows operators to expand capacity in phases rather than committing to a full build-out upfront - a useful feature when AI compute demand can shift quickly.
Why this matters for IT and development professionals
Modular data centers change the calculus for capacity planning. When infrastructure can be deployed in factory-built blocks, the gap between ordering compute and powering it on shrinks. For teams managing AI training or inference workloads, that means less waiting on physical build-outs and more flexibility to scale with project needs. The retrofit model also opens possibilities for repurposing existing facilities, which could bring compute closer to where data already lives rather than requiring greenfield construction in remote locations.