Concerns are rising that the investment and construction of data centers, which are key infrastructure in the era of artificial intelligence (AI), are concentrated in the United States and China, potentially widening the AI gap between G2 countries and others. Voices are growing that Korea needs to create an environment where AI data centers can be established stably in order to keep pace in the AI competition.
According to industry sources on the 11th, OpenAI stated in a blog post titled "Why data center investments are concentrated in the United States and China" published the previous day that "AI infrastructure is a decisive factor influencing national competitiveness, but investment plans for AI data centers are still focused on the United States and China," noting that "unless there is a change in energy prices and data accessibility, this trend will continue."
According to the Financial Times (FT), China has completed or is constructing over 250 AI data centers under government initiative. The United States is expanding large AI data centers led by big tech companies such as OpenAI, parent company Meta of Facebook, Amazon, and xAI, led by Elon Musk, CEO of Tesla.
Energy availability, which refers to the ability to receive stable power when needed, and competitive energy prices are cited as the main factors affecting AI infrastructure investments. OpenAI pointed out that despite the European Union (EU) announcing plans for investment in large data centers called "AI giga factories," the high electricity rates make them less attractive compared to the United States and China. Currently, the average industrial electricity rate in the EU is three times that of the United States and four times that of the United Kingdom. AI data centers consume tremendous electricity, with energy usage per unit area being 10 to 20 times higher than that of general data centers.
OpenAI noted that copyright and data-related regulations are barriers that lower the possibility of the EU becoming an AI hub. It remarked that the United States' "fair use" principle has historically been a driving force for technological innovation and that the AI industry has been able to grow rapidly thanks to this principle. OpenAI expressed that "Japan and Singapore have also established copyright protection systems similar to the U.S. fair use for innovative corporations" and remarked that "if they have only energy price competitiveness, it could serve as a positive factor for attracting future AI investments."
As energy prices and data regulations emerge as key variables in attracting investments in AI data centers, claims are made that Korea must enhance its related infrastructure competitiveness to avoid falling behind in the global AI competition.
Korea has set a goal to leap forward as one of the top three AI powers, but it appears that the current power infrastructure will struggle to support AI data centers. Experts point out that the increased instability in power supply and electricity rates in recent years have undermined private investment incentives. According to the Korea Chamber of Commerce and Industry, as of the end of last year, Korea's industrial electricity rate was 190.4 won per kilowatt-hour (kWh), more than 40% higher than that of China (129.4 won) and the United States (121.5 won), and similar to that of France (197.1 won).
Ryu Yong, head of the Infrastructure Strategy Center of the Korea Telecommunications Operators Association (KTOA), stated, "AI data centers consume so much power that they are referred to as 'electricity-eating hippos,'" and added that "dedicated electricity rates for AI data centers should be established to lessen the burden on corporations."
In Korea, over 60% of data center infrastructure is concentrated in the metropolitan area, and there are analyses that the excessive concentration has already led to power consumption in the metropolitan area reaching its limits, putting a burden on the power grid. According to the Ministry of Trade, Industry and Energy, the domestic electricity usage requests for data centers are projected to reach 7,343 megawatts (MW) by 2027, while the supply capacity is only 4,718 MW. This indicates that the power shortage for data centers is imminent.
Industry insiders argue that to respond to the rapidly increasing demand for AI data centers, the government should actively encourage regional distribution and expand power grid infrastructure to prevent overload. The government has stated it will build an "energy highway," and the Korea Electric Power Corporation (KEPCO) has announced a power grid investment plan worth 73 trillion won. However, securing the necessary funds remains a challenge, given that KEPCO is currently facing financial difficulties with a cumulative deficit of 40 trillion won.
Domestic copyright-related regulations have also been identified as factors hindering AI learning and development. Ryu noted, "Currently, even data publicly available on the internet requires prior consent, and there is a need for structural improvements such as establishing a public data hub and shifting to a judgment of infringement centered on generated outcomes."