SK Telecom, Supermicro & Schneider Partner on AI Data Centers

The collaboration hinges on a "pre-fabricated modular" approach, integrating servers, power, and cooling into single modules before shipping. This method significantly differs from traditional data center construction, where a building is completed first, followed by the sequential installation of servers and other infrastructure. The goal is to slash deployment times and navigate around supply chain bottlenecks that can stall projects for months. At the core of this partnership is the challenge of cooling high-density AI hardware. As AI workloads intensify, traditional air cooling is hitting its thermal limits, unable to dissipate the heat generated by dense clusters of GPUs. Liquid cooling, which can be up to 90% more energy-efficient than air-based systems, is becoming a necessity for modern AI data centers. Schneider Electric will provide this critical mechanical, electrical, and plumbing (MEP) infrastructure, including advanced liquid cooling solutions. Supermicro's role is to supply the high-performance GPU servers optimized for specific AI computing scenarios. The company specializes in creating server and GPU clusters for demanding AI training and inference tasks and has experience building large-scale AI infrastructure with advanced liquid cooling. This includes everything from NVIDIA HGX-based systems to compact servers for edge AI deployments. SK Telecom brings its extensive experience in operating large-scale data centers to the table. The partnership allows for a more scalable and flexible approach to building capacity. Instead of massive upfront investments, modules can be deployed in phases as demand for AI services grows, offering a more cost-effective and adaptable model for hyperscalers and other large customers. This addresses a key industry challenge, as the demand for AI is expected to drive 13% to 20% annual growth in global IT power capacity through 2030.