Samsung Electronics signed an agreement to build a 240 billion won, approximately $170 million, HVAC production line in Gwangju, South Korea, marking the largest manufacturing commitment at that campus in 37 years. The Samsung CDU factory AI cooling demand surge is the underlying driver. The deal plants Samsung’s FlaktGroup cooling subsidiary in the country where demand for the underlying hardware is climbing fastest. At the center of the bet sits the Coolant Distribution Unit, the unglamorous box that makes liquid-cooled AI data centers thermally possible. The agreement, signed at Gwangju City Hall with FlaktGroup Korea and provincial officials, repositions a former appliance campus into Samsung’s global anchor factory for CDU production.
A Coolant Distribution Unit is not a conventional air conditioner and does not cool an entire data hall. Instead, it cools individual chips. In direct-to-chip liquid cooling architectures, the CDU sits between a facility’s main chilled water supply and the IT hardware’s own closed loop, the network of tubes and cold plates bolted directly onto each GPU and CPU package. This separation is not incidental because facility water carries dissolved minerals, biological organisms, and chemicals that would corrode the precision copper cold plates mounted on billion-dollar processors.
The CDU creates a thermally isolated secondary loop, filtering and conditioning a separate water-glycol mixture that circulates through the IT equipment, while a heat exchanger transfers chip heat outward to the facility’s chilled water without the two fluids ever mixing. Inside that loop, a cold plate, a flat metal block with micro-channels machined through it, mounts directly onto the processor package. Coolant flows through those micro-channels at a controlled flow rate and temperature setpoint, absorbing heat through direct conduction at the chip surface, then returning to the CDU to release heat, chill back down, and recirculate.
The reason this architecture has become mandatory is physics. NVIDIA’s Blackwell B200 GPU draws roughly 1,200 watts in liquid-cooled configurations, about 71 percent more than the H100 SXM’s 700 watts. Packed into the GB200 NVL72 chassis, which holds 72 GPUs and 36 Grace CPUs in a single rack, that translates to 120 to 132 kilowatts per rack at full load. Air cooling works adequately below about 30 to 50 kW per rack, while the GB200 NVL72 requires roughly three to four times that. Liquid carries approximately 3,500 times more heat per unit volume than air at equivalent flow rates, making liquid cooling not an option but a requirement.
The Gwangju campus has manufactured Samsung consumer appliances, including Bespoke refrigerators, washers, dryers, and air conditioners, since Samsung established the site in 1989. The new HVAC production line spans 21,800 square meters at Campus 3 in Buk-gu, the first major production facility investment at the site in those 37 years. That distinction signals a board-level decision to reposition an existing campus around a new industrial priority, one Samsung expects to grow faster than its legacy appliance business.
At 240 billion won, the Gwangju commitment dwarfs the roughly 5.7 million euro investment Samsung and FlaktGroup made in a Pune, India facility that opened on August 12. India’s Pune plant produces up to 6,500 HVAC units annually and was the first production line under the FlaktGroup acquisition. Gwangju represents something different in kind: a manufacturing hub investment at Samsung’s home base, designed to become FlaktGroup’s 15th global production site and join an existing network serving customers in 65 countries.
South Korea is arguably the most compressed AI infrastructure buildout in the world. The national government has committed to 8.4 gigawatts of AI capacity by 2029, backed by roughly 550 trillion won in public and private investment. A separate AIDC Alliance is targeting 18.4 GW by 2035, supported by 200 trillion won in state financing. Samsung Electronics itself committed to group-wide investment of 2,450 trillion won from 2026 to 2040, with a semiconductor cluster in Gwangju as its anchor.
Jeollanam-do Province, where Gwangju sits, is running the fastest capacity growth in the country at a 23.05 percent compound annual growth rate on new data center additions. The overall South Korean data center construction market was valued at approximately $7 billion in 2026 and is projected to reach $14.6 billion by 2031. A domestic CDU manufacturer operational by early 2028, the planned ramp date, means Korean operators building AI campuses now can plan cooling procurement with shorter lead times and without importing from European or American competitors.
Cheolgi Kim, Executive Vice President and Head of Digital Appliances at Samsung Electronics, framed the Gwangju commitment as a strategic expansion. FlaktGroup CEO David Dorney added the geographic calculus, noting that combining Samsung’s AI expertise with FlaktGroup’s HVAC product capabilities positions the combined company as an HVAC partner of choice in Korea and across Asia. The new factory is explicitly described as a springboard for full-scale expansion into the Korean market, with AI data centers foremost among targeted sectors.
The investment’s deepest significance is not the factory itself but what the factory completes. Samsung’s FlaktGroup acquisition brought CDU manufacturing capability into the company. Its semiconductor division produces HBM4, the high-bandwidth memory that populates every leading AI accelerator. Combined with Samsung’s broader memory and chip operations, the Gwangju plant extends a vertically integrated stack that few competitors can replicate, spanning silicon, packaging, and now the thermal infrastructure required to keep advanced accelerators running within spec.
Full operations are scheduled to begin in early 2028, a timeline the company says aligns with the wave of new AI campuses completing construction across South Korea. As liquid-cooled rack densities climb toward the 250 to 900 kW range projected for NVIDIA’s Rubin generation, the CDU will move from a niche component to a foundational building block of AI infrastructure. Samsung’s Gwangju bet reads less like a manufacturing expansion and more like a positioning move for the next decade of compute growth.
The combination of Samsung’s semiconductor memory leadership with its newly anchored CDU manufacturing in Gwangju creates a uniquely integrated response to rising Samsung CDU factory AI cooling demand across the Korean peninsula and the wider Asian market. No other vendor currently pairs high-bandwidth memory production with domestic coolant distribution hardware at this scale, and the 2028 ramp will test whether that integration translates into durable market share.