LG CNS rolls out container-based AI Box for data centers

LG CNS has rolled out a container-based AI Box that holds up to 576 GPUs and can be deployed in about six months.

Each module can run standalone or link in clusters of dozens to form hyperscale AI data centers, letting customers start small and scale while keeping upfront investment low.

Each box supports about 1.2 MW of IT load and has an electrical room for UPS, transformers, and power distribution, plus a computing room for servers, GPUs, and cooling.

LG CNS said the design integrates power, cooling, and IT systems and uses cooling gear from LG Electronics and battery systems from LG Energy Solution that support UPS.

The company plans to deploy the first unit at its global cloud data center site in Busan and about 50 units across a 27,179-square-meter campus.

LG CNS said the AI Box offers rapid deployment and flexible scaling, an alternative to large data centers that often face delays from site acquisition, permits, and complex cooling and power design.

🔗 Source: The Korea Times

🧠 Food for thought

Implications, context, and why it matters.

The “One LG” strategy underscores the AI Box’s integrated design

LG CNS casts the AI Box as a modular data center “package” that bundles power, cooling, plus IT systems, using LG Group affiliates through a “One LG” setup 1.
It uses cooling gear from LG Electronics plus battery systems from LG Energy Solution for uninterruptible power supply (UPS) 2.
The “One LG” setup pulls together servers and GPUs, plus power, cooling, plus operations, yet sources do not back claims of better hardware or software optimization than competitors 2.
Plans for about 50 units at LG CNS’s global cloud data center site in Busan put total IT load near 60 megawatts (50 x 1.2 MW), matching reported figures of about 50 modules plus 1.2 MW per AI Box 3.

This reframes AI infrastructure deployment as a standardized, modular offering

The build timeline drops to roughly six months, versus more than two years for traditional data centers, which changes how fast companies add AI data center capacity 1.
LG CNS says enterprise and cloud data center customers can start with one container, then grow to dozens, while easing upfront spending and tracking demand 4.
This model treats capacity as repeatable modules, rather than custom builds, so operators can react faster to market shifts 1.
LG CNS has said it plans to move into Southeast Asia plus North America, using domestic rollouts as a springboard 2.