AI factories become a headache: Data centers strain resources

Data centers, initially designed to house early, room-sized computers like Britain's 'Colossus' and the U.S.'s 'ENIAC,' have transformed dramatically. They have evolved from mere storage spaces for government-owned machines to central hubs for private companies' server rooms. Today, these facilities are increasingly becoming 'AI factories,' serving as critical bases for training, running, and generating data for artificial intelligence models.

Despite technological advancements that have miniaturized computers, a fundamental characteristic of data centers remains unchanged: their immense appetite for electricity and water. This is driven by the need to power high-performance semiconductor chips and dissipate the significant heat they generate. Consequently, the scale of data centers is often measured not by physical size or server count, but by their power capacity in megawatts (MW) or gigawatts (GW).

In the United States, the impact of this demand is palpable. The planned power capacity for the top 90 data centers under construction exceeds 205 gigawatts, representing about 15% of the total U.S. power generation capacity. This surge has led to skyrocketing electricity and water prices, alongside noise pollution from cooling and power generation equipment, causing distress among residents in affected areas. In response, New York state has taken the unprecedented step of imposing a one-year moratorium on the construction of new large data centers.

While the U.S. grapples with the downsides, South Korea is actively pursuing data center development. The government plans to build 18.4 gigawatts of AI data center capacity by 2035, requiring an area equivalent to 6.3 times that of Seoul's Yeouido district. This push is further supported by recent legislation aimed at easing regulations for data center construction in local areas. The contrasting approaches raise questions about the future balance between technological advancement and local community well-being.