[Edaily Reporter JAEMIN SONG ] As the number of artificial intelligence (AI) data centers grows rapidly, the cooling market—which cools the heat generated by servers—is expanding alongside it. In particular, high-performance AI chips consume more power and generate more heat than conventional server semiconductors, making it increasingly difficult to rely solely on traditional air-cooling methods. As a result, liquid cooling—which uses water or coolant to directly cool servers—is emerging as a new alternative, and SamsungElectronics and LGELECTRONICS are accelerating their market expansion through production facility expansions and large-scale supply contracts.

According to industry sources on the 6th, Goldman Sachs raised its forecast for U.S. data center installed capacity by the end of this year to 64 gigawatts (GW)—an increase of 5 GW from its previous estimate—in a report released on the 4th (local time). Analysts note that despite local opposition to data center construction and the burden of electricity costs, the actual pace of project development is faster than expected. Separately, the firm projected that power demand from U.S. data centers will increase by 38% year-over-year, or 12 GW, this year.

As the number of data centers increases and the performance and integration density of AI servers improve, the importance of cooling systems is growing even more. According to market research firm TrendForce, the adoption rate of liquid cooling for AI chips is expected to rise from approximately 33% last year to 53% this year and reach about 60% next year. This is because the power consumption of the latest AI chips exceeds 1 kilowatt (kW) per chip, and power consumption per rack is rising to several hundred kW, making it increasingly difficult to cool servers using air alone.

Cooling methods are also shifting from lowering the temperature of the entire data center space to a approach that brings cooling closer to the servers and chips where heat is generated. Liquid cooling uses coolant to directly remove heat generated by servers, enabling it to handle higher thermal densities than air cooling. As global AI chip manufacturers, such as NVIDIA, expand the use of liquid cooling in their next-generation platforms, demand for related equipment—including large chillers and cooling water distribution units (CDUs) that supply and circulate coolant to servers—is also expected to rise.

Domestic companies are also expanding their cooling businesses. LGELECTRONICS announced yesterday that it had signed a long-term contract with Air Control Concepts (AIR), a North American infrastructure platform company, to supply cooling solutions for AI data centers totaling 5 GW in capacity. The contract value and specific supply period were not disclosed. The companies are discussing a plan to first supply high-efficiency chillers in phases and subsequently expand the scope of supply to include liquid cooling solutions such as CDUs.

According to LGELECTRONICS, the value of orders for AI data center cooling solutions exceeded 600 billion won in the first half of this year alone. The company expects cumulative orders to reach several trillion won by the end of the year. LGELECTRONICS is pursuing a “chip-to-chiller” strategy that encompasses everything from chillers to CDUs, which circulate cooling water close to the servers. Recently, the company expanded its high-capacity product lineup by securing “DSX Ready” certification—NVIDIA’s data center design standard—for its 2.5 MW CDU, in addition to its existing 600 kW and 1 MW models.

SamsungElectronics is also expanding its production base. The company is investing approximately 240 billion won at its Gwangju plant to establish a heating, ventilation, and air conditioning (HVAC) production line for Flact Group, a global HVAC specialist it acquired last year. Scheduled to begin operations in early 2028, this facility will produce CDUs, fan wall units (FWUs), and computer room air handling units (CRAHs) for AI data centers. The plan is to combine Flact Group’s cooling technology with SamsungElectronics’ manufacturing capabilities to meet data center demand in South Korea and East Asia.

An industry official stated, “As the performance of AI servers increases, power consumption and heat generation issues are bound to grow,” adding, “With the expansion of data center investments, cooling equipment will become an even more critical component of AI infrastructure.”