HWNEWS
South Korea's September energy plan faces a timing problem: AI and data center demand arrives in 2 4 years, new reactors take 9 10, and renewables are not at scale until 2030.
Korea's ruling Democratic Party is weighing whether to build new nuclear reactors to power the Lee Jae Myung administration's three flagship industrial projects: semiconductors, data centers, and physical AI. Climate, Energy and Environment Minister Kim Sung-whan said on July 13 he is not ruling out new nuclear, and the country's fifteen-year energy roadmap, the 12th Basic Plan for Electricity Supply and Demand, due in September, will set the terms.
The two fuel paths share the same fifteen-year clock, and that is the bind. Korea Hydro & Nuclear Power, the state operator, builds APR1400 reactors on a nine-to-ten-year schedule. The renewable target is 100 gigawatts by 2030, up from 34 GW at the end of 2025. AI and data-center demand is on a two-to-four-year curve. Storage is supposed to firm the gap, and the plan names 3.7 GW of energy storage by 2030 as a floor. The minister's argument, as reported, is that storage alone is not enough to ride out the intermittency while the renewable build-out is still scaling.
Korea JoongAng Daily carried the July 13 comments alongside a more specific picture: the party wants firm power for the industrial build-out, and the trade press has already matched the signal. NucNet, the nuclear-industry news service, described the plan as a way to power the same flagship projects.
The two candidate sites Kim named on July 3 reflect the gap, not a finished answer: two more reactors at the Hanbit complex in Yeonggwang, South Jeolla, and two at the Ulju site in Ulsan. Both are expansions of existing plants, which trims some licensing and grid work. KHNP has already shortlisted sites for both large reactors and small modular reactors, the industry shorthand for factory-built units under 300 megawatts that can be sited in pairs instead of as single megaprojects.
The case for reopening the nuclear question is not purely a clean-energy story. AI training and inference are the load that justifies new build; the same load is also what is pulling Korean utilities to sign long-term contracts with industrial data-center tenants at prices the household sector has not seen. The construction side is no less real. APR1400s are not cheap, and the country's last build program ran over budget and behind schedule. Restarting reactors adds a separate political economy: who pays for the deferred maintenance, what happens to spent fuel on already crowded sites, and how a grid operator weighs the variable output of new solar and wind against a 1.4-gigawatt unit that runs day and night.
The September plan will be judged on whether it names the bridge load for the gap years and pays for it on terms that do not cannibalize either fuel. The test is concrete: does the document identify what runs in 2027 and 2028, when AI demand is breaking ground, the new reactors have not yet broken ground, and the renewable build is still on its way to 100 GW? If the answer is restarts, lifetime extensions of older reactors, transmission build-out, demand response contracts with the chipmakers, or a faster small-reactor program, the plan will read as a real engineering document. If the answer is only a list of reactor names and a 2030 target, it will read as a press release.
Presidential chief of staff Kang Hoon-sik has said the government will look at ways to shorten the APR1400 build time. The editorial board of Chosun.com, the English edition of the Chosun Ilbo, has framed the choice as a security one: the country cannot afford to be off either fuel. That is a position, not a forecast. The plan itself, when it lands, is the test.