AI Infrastructure Race Shifts From Expansion to Cost Efficiency

Cost reduction in data center construction has emerged as the top priority in the artificial intelligence industry. While companies have focused solely on expanding AI infrastructure, the ability to operate sustainably is becoming increasingly critical.

SK Telecom announced Wednesday that it signed a three-way memorandum of understanding with U.S. server manufacturer Supermicro and France's Schneider Electric at Mobile World Congress 2026 in Barcelona, Spain. The partnership aims to secure integrated solutions using "prefab modular" construction methods, focusing on shortening AI data center construction timelines.

The prefab modular approach involves pre-manufacturing power, cooling, and IT infrastructure in modular units before assembling them on-site, dramatically reducing AIDC construction periods. Unlike conventional methods where servers are sequentially installed after building completion, this approach integrates AI computing servers with power and cooling infrastructure into single modules, improving both construction speed and cost efficiency.

SK Telecom also signed an MOU with Panesia, a domestic fabless startup. Rather than simply adding more graphics processing units in AIDCs, the partnership seeks to improve performance and cost efficiency by changing how computing resources are connected. Panesia will provide telecommunications semiconductors needed to build efficient AIDCs. The two companies plan to transition from fixed server-unit structures for CPUs, GPUs, and memory to flexible connection architectures. This expands resource connectivity from within individual servers to rack-level clusters of multiple servers, enabling selective utilization of needed resources.

Other telecommunications carriers are also presenting solutions to reduce AIDC operating costs. Japan's NTT showcased its vision at its booth to cut data center power consumption to one-hundredth of current levels by 2032. IOWN, the technology to achieve this goal, is a next-generation optical communications network that maximizes data processing capacity and dramatically reduces power consumption by using light instead of electromagnetic waves.

NTT also presented solutions to optimize data center operations. When large-scale data processing is required, data centers distributed across multiple regions operate collectively; when data volumes decrease, some data centers are deactivated. An employee at the booth explained, "NTT has achieved ultra-low latency in data processing through optical communications technology. This allows us to operate multiple data centers across Europe as if they were a single data center."