SK Telecom said on the 4th that at MWC26, the world's largest mobile communications exhibition being held in Barcelona, Spain, it signed a three-party memorandum of understanding (MOU) with global server manufacturer Supermicro and Schneider Electric SE, a global manufacturer in the mechanical, electrical and plumbing (MEP) field for artificial intelligence (AI) data centers (DC), to secure an integrated solution using a pre-fabricated modular approach. The company said it will pursue a new build model that reduces the time and expense required to build AI DCs through this agreement.
The core of the collaboration is the pre-fabricated modular method, which pre-manufactures power, cooling and IT infrastructure in module units and assembles them on-site. The company said the biggest advantage is that it can drastically shorten the period required to build AI DCs. Unlike the existing steel reinforced concrete (SRC) method, in which servers are sequentially installed after building completion, this approach integrates the servers that handle AI computation and the supporting power and cooling infrastructure into a single module for integrated production, simultaneously increasing build speed and expense efficiency. In addition, modules can be added step by step as demand grows, allowing flexible responses to rapidly changing market demand while reducing the burden of large upfront investment.
Supermicro is a global manufacturer of server and storage systems. Based on close cooperation with major AI Semiconductor corporations, it rapidly designs and supplies high-performance GPU servers specialized for AI training and inference. In particular, last year it worked with major U.S. big tech and AI infrastructure corporations to build a large-scale AI factory that applies servers based on the Blackwell GPU and advanced liquid cooling technology. In this collaboration, Supermicro plans to build high-performance servers that perform AI computation and GPU clusters that efficiently bind them, so as to deliver optimal performance tailored to customer requirements and use scenarios. Supermicro will prepare an integrated configuration that takes into account key infrastructure elements such as cooling, enabling faster construction of DCs optimized for high-performance AI computation.
Schneider Electric SE is a global corporations in energy management and automation, with solutions that have become global standards in DC power management and digital-based operations management. In particular, it ranked No. 1 in the world for the second consecutive year on Time's list of the World's Most Sustainable Companies released last year in the United States. In this collaboration, Schneider Electric SE plans to present an MEP-based integrated AI DC model that considers everything from infrastructure design to operational efficiency to reliably meet large-scale AI demand.
Ha Min-yong, who oversees SK Telecom's AI DC business, said, "We have come to pursue an integrated solution using a pre-fabricated modular approach by working with global partners representing the AI DC field," adding, "Based on this, we will proactively respond to global big tech customers' AI DC build demand while also enhancing competitiveness in terms of expense."
Supermicro Vice President Clay Simmons said, "We are very pleased that cooperation with SK Telecom allows us to explore an integrated framework that can further accelerate DC construction and operation," adding, "This integrated solution will be based on Supermicro's high-performance GPU-specialized servers optimized for customer workloads." He added, "We will continue to provide the latest technologies to respond to surging DC demand."
Andrew Bradner, senior vice president at Schneider Electric SE, said, "In the AI era, competitiveness hinges on how quickly and sustainably you can build high-performance infrastructure," adding, "With this collaboration, we will present a pre-fabricated modular-based integrated AI DC model to reduce carbon emissions and ease supply bottlenecks, while also reliably supporting customers' high-density AI workload operations."