Regions across Korea are competing intensely to attract AI data centers. South Chungcheong Province has declared itself an 'AI capital' after attracting successive AI data center projects worth around 3 trillion won in Dangjin alone, while Saemangeum and Gangneung have also proposed projects worth trillions of won. Yet the situation in the United States is quite different. Last year, 48 data center projects worth $156 billion were halted or delayed by opposition from local residents. Residents concerned about higher electricity bills and water shortages blocked permits, and the issue became a point of contention in the Texas midterm elections. The same facilities bring hope in some places and disputes in others. The reasons for that difference deserve examination.
AI data centers consume electricity, water and land. A single 100-megawatt facility uses as much electricity as a city and draws enormous quantities of cooling water to cool its servers. There are even forecasts that global data center electricity consumption will exceed South Korea's entire consumption next year. The problem is that these burdens are not readily visible. Investment and employment are expressed in numbers, but nobody can understand how much substation capacity remains, how cooling water affects rivers during heat waves, or how much server heat warms nearby residential areas until those factors are examined together on a map. In the author's view, the opposition in the United States is the consequence of failing to anticipate these less visible burdens. Competition to attract investment focuses on being first, but location is ultimately what matters.
Determining that location involves more complex calculations than might be expected. Factors include spare grid capacity, available water withdrawals, ground conditions and flood risks, the extent of heat-island effects, distance from residential areas, and transmission line routes. Each condition may appear acceptable when examined separately, but considering them together can reveal that a site is unsuitable. It is difficult for people to track all these overlapping conditions by eye. This is where spatial intelligence, or geospatial AI, is needed. Before an agreement is signed, simulations can determine how many years it would take for a substation to reach its load limit if a site receives 100 megawatts, or how much river temperatures would rise when summer electricity peaks coincide with cooling-water demand. Site assessment should be completed before investment is attracted, rather than follow it. Attention should therefore turn to which cities can perform these assessments and whether those cities themselves have AI data centers.
Daejeon has so far been quiet in the competition to attract these facilities. But an AI data center is more than a building full of servers. It collects regional transport, safety and welfare data to run AI services, provides the computing foundation for a simulation city in which robots and autonomous vehicles are tested tens of thousands of times before entering real streets, and supplies the 'AI infrastructure' required by national funding competitions. Without it, even technology developed in the Daedeok Innopolis will run on servers in other regions. A pattern would become established in which research is conducted in Daejeon but its results are used elsewhere. Daejeon should therefore actively seek to attract AI data centers. Fortunately, its site conditions are competitive. It has abundant water resources in the Daecheong Lake system, and nearby institutions responsible for power grids, water resources and weather could help it complete those complex assessments ahead of others. Daejeon's approach should be to use data to identify sites with suitable electricity and water supplies, present candidate locations first, and work with South Chungcheong Province and Sejong to plan the entire Chungcheong region's power grid and water systems on the same map. The region that locates AI data centers most effectively, rather than attracting the largest facilities or the greatest number, will ultimately become a model for local governments across Korea and cities overseas.
AI data centers are long-term industrial facilities tied to their sites. Once built, they draw electricity and water from the same location for decades. Their location shapes a region's next 10 years. Daejeon does need AI data centers, both to prevent research activity from moving elsewhere to access servers and to demonstrate that a city best equipped to assess suitable locations can put that knowledge into practice. Daejeon can take the lead in shifting the competition from attracting investment to selecting appropriate sites.
Original Korean article: daejonilbo.com
English brief by Korea Wind Intelligence, koreawindintel.com
