Korea Research Institute of Chemical Technology, SK Innovation team up to tackle AI data center heat problem

South Korea's state-run Korea Research Institute of Chemical Technology and SK Innovation have joined forces to develop next-generation cooling technology for AI data centers, which have drawn growing criticism for their strong power consumption.

The Korea Research Institute of Chemical Technology said Wednesday it signed a memorandum of understanding with SK Innovation's Environment Science & Technology Institute to cooperate on materials and technology research related to energy solutions.

As AI adoption drives a surge in data centers running large numbers of high-performance GPUs, the ability to remove server heat quickly and efficiently has become a key competitive factor. Conventional air-cooling systems are increasingly unable to handle the thermal load generated by high-density servers.

The technology both institutions are focusing on is two-phase cooling, in which a liquid refrigerant absorbs heat from servers, vaporizes, and then condenses back into liquid — transferring large amounts of heat in the process. Securing refrigerants that absorb heat rapidly while remaining stable over extended use, along with effective heat-dissipation and heat-transfer materials, is central to improving cooling efficiency.

The two organizations plan to combine AI and chemical materials expertise to identify new two-phase refrigerants. The approach involves using machine learning to predict the thermal and physical properties of candidate substances, selecting the most promising ones, then synthesizing and refining them and verifying key properties such as thermal conductivity and latent heat.

Large-scale demonstration is also on the agenda. Under a research plan currently under review, the collaboration will begin with AI-driven refrigerant discovery, move through synthesis and refining process optimization and performance verification at the server and rack level, and culminate in validating cooling performance in a 500kW-class AI data center environment.

The partners also plan to confirm power usage effectiveness — the key power-efficiency metric for data centers — as well as long-term operational reliability.

Alongside this, the two institutions will jointly identify national research and development projects in AI data centers and electrification, and will cooperate on developing high-efficiency, eco-friendly materials for heat management, cooling, heat dissipation and power infrastructure. They also plan to expand exchanges of research personnel and technical information.

"By combining the institute's research capabilities in refrigerants and thermal-management materials with SK Innovation's experience in energy research and development, we aim to secure core source technologies in advanced materials that will transform power efficiency in AI data centers," said Shin Seok-min, president of the Korea Research Institute of Chemical Technology.

Kim Pil-seok, head of SK Innovation's Environment Science & Technology Institute, said responding to the rapid growth of AI data centers requires integrated technological innovation spanning refrigerants, materials and cooling systems. "We will work to proactively secure next-generation energy solution technologies and apply them in industrial settings," he said.