Korea's 720-billion-dollar AI plan meets its first critic: the man running its research hub

Korea has designated AI data centers a national strategic industry, after semiconductors, and published a blueprint calling for more than 1,000 trillion won ($720 billion) of investment through 2035 to make it one of the world's top three AI powers ^[1]. The man appointed to run the country's new national research anchor has used an interview to argue that the problem with the plan is not the size of the number but the calendar it sits on ^[2][3].

Kim Kee-eung, inaugural director of the National AI Research Hub and a professor at KAIST's Kim Jaechul Graduate School of AI, told the Seoul Economic Daily on the 17th that AI requires "setting and executing budgets and investment plans with an eye on the nation's future over more than a decade, not one or two years" ^[2][4]. Nothing, he said, is more important to building an AI ecosystem than creating patient capital of a long-term investment nature ^[3].

The arithmetic explains the anxiety. Spread evenly across the ten years ending 2035, $720 billion is about $72 billion a year ^[1]. Kim's own description of the United States has Big Tech investing hundreds of billions of dollars annually to build ecosystems including data centers, on top of an enormous pool of investment funds that already functions as patient capital ^[5]. On those figures, Korea's decade-long national commitment is smaller than one year of American private spending, and the thing Korea has to construct is the thing the US market supplies as a matter of course. The headline is also a won number: it implies roughly 1,389 won to the dollar ^[2], so the dollar value of the pledge moves with the currency, not with policy resolve.

The second constraint is people. From 2022 to 2025 Korea was a net exporter of AI talent, according to Kim, and he wants extraordinary terms and lower barriers to bring overseas researchers back ^[6][7]. The hub itself launched in October 2024 with support from the Ministry of Science and ICT ^[8], inside that window of outflow ^[3]. It spans KAIST, Korea University, Yonsei and POSTECH, with 45 faculty members, some 200 student researchers, 12 domestic partner companies and 14 overseas joint research institutions ^[8] - about 245 researchers in total ^[4]. It was organised in line with the US National AI Research Institutes, the UK's Alan Turing Institute and Canada's Vector Institute ^[9].

Kim's framing of the competition is that the US now treats AI as infrastructure running through science, energy and defence rather than an industrial technology ^[10], citing a July AI Action Plan on regulation, infrastructure and full-stack AI exports and a November "Genesis Mission" aimed at doubling scientific research productivity within a decade ^[11]. China, he said, set a 2017 goal of world leadership by 2030, then moved to an "AI Plus" action plan applying AI across six core industries including manufacturing and healthcare, with an open model strategy following DeepSeek ^[12][13]. He describes US export controls on advanced chips as blocking Chinese computing access outright ^[14].

Watch the appropriation mechanism rather than the target: whether the 1,000 trillion won arrives as multi-year committed tranches or annual budget lines that can be trimmed, and whether spending flows to data center capex or to the researcher base, which currently numbers in the hundreds ^[8][4]. A reversal in the talent balance after 2025 would be the first honest signal that the terms changed ^[6].