South Korea's Nuclear Debate Puts Electricity Demand Forecasts Under Scrutiny

A forecast that could shape the nuclear decision

South Korea's public debate over nuclear power will ask how much of the country's future electricity should come from reactors. Critics say an earlier question needs equal attention: how much electricity will the country actually need?

At its second meeting on October 14, 2026, the Nuclear Power Public Deliberation Committee set the agenda for discussions feeding into the 12th Basic Plan for Electricity Supply and Demand. Its main question is what kind of electricity South Koreans will live with in the future, with a narrower focus on nuclear power's role and share in meeting future demand.

The stakes have grown as semiconductor factories and artificial intelligence projects have entered the government's calculations. The revised projection puts peak electricity demand in 2040 at 158.4 to 165 gigawatts, up by as much as 26.8 gigawatts from an April estimate. That increase is roughly equal to the generating capacity of 19 APR1400 nuclear reactors.

Lee Heon-seok, a policy committee member at Energy Justice Actions, argues that treating the demand forecast as fixed could steer participants toward supporting more reactors before they have examined whether the projected consumption will materialize. Seok Gwang-hoon, an expert committee member at the Energy Transition Forum, warns that predictions of shortages could put psychological pressure on citizen representatives.

The dispute now reaches beyond campaign groups. Climate, Energy and Environment Minister Kim Sung-hwan acknowledged during an October parliamentary audit that data center demand had been overestimated in some respects. The decision will affect spending on power plants, transmission lines and, eventually, electricity bills.

What citizens will decide

The committee met at the Electricity Infrastructure Center in Seoul's Yeongdeungpo District. It excluded four subjects from deliberation: extending the operation of individual reactors, nuclear projects whose sites have already been selected, the government's announced renewable energy deployment targets, and the selection of specific project locations.

Reactor operating extensions were excluded because they fall under safety reviews by the Nuclear Safety and Security Commission. The remaining discussion will concern nuclear power's place in the electricity system rather than a vote on a particular construction site.

The process is designed to last three months. A citizen panel will receive information, deliberate and answer surveys before and after discussions. Expert debates and online public feedback will accompany its work. An oversight committee of up to 11 members is intended to draw on fields including statistics, conflict management, science and the humanities, with nuclear specialists providing advice separately.

A supporting expert group of no more than 10 people from environmental, nuclear and neutral fields will prepare a briefing book and online videos. Environmental and nuclear organizations will be asked to recommend specialists. Detailed discussion topics and survey questions are to be developed from the committee's next meeting.

Committee chair Eun Jae-ho said he hoped the process would help move nuclear policy beyond partisan and ideological divisions. Its design faces objections from both directions. Critics of expansion question the demand assumptions, while Dankook University energy engineering professor Moon Joo-hyun has questioned why renewable expansion is excluded from similar deliberation and warned that the government could shift responsibility for its decision to citizens.

How the demand forecast is built

The Ministry of Climate, Energy and Environment's draft demand outlook, discussed at the 38th Energy Commission meeting, describes a calculation with several layers. A model first uses economic output, population and perceived temperature. Separate estimates then add electricity for advanced industries, data centers and electrification, meaning the replacement of fuels such as oil and gas with electricity.

The ministry then subtracts expected savings from efficiency measures and adjusts for consumption shifted away from peak periods. The resulting target demand helps determine how much generating capacity and grid infrastructure the country should build.

The preliminary outlook listed additional advanced industry consumption of 29.3 terawatt hours in the base scenario and 27.4 terawatt hours in the higher scenario. The ministry explained that the amount added separately depends on how much industrial consumption is already captured by the economic model. It also deferred some longer term semiconductor demand for consideration in later plans as business conditions become clearer.

Data centers added 26.5 terawatt hours, calculated using projected CPU and GPU server numbers and assumptions about cooling efficiency. Electrification added 112.6 to 119.4 terawatt hours. Expected demand management savings were similarly large, at 123.2 to 125.6 terawatt hours.

The assumptions reach into everyday life and heavy industry. They include electric vehicles, heat pumps for buildings, electric steelmaking and electricity used to produce hydrogen. The transport calculation extends a deployment trend based on 8.88 million electric vehicles in 2035.

Annual consumption and peak demand answer different questions. Terawatt hours measure electricity used over time; gigawatts measure the power needed at a particular moment. A country must supply enough electricity across the year while also meeting its busiest hours. Renewable capacity, reactor capacity and annual electricity consumption therefore enter different parts of the planning calculation.

Why the numbers changed so quickly

The government's three megaprojects cover semiconductors, physical AI and AI data centers. Physical AI refers to systems that act in the real world, including robots and automated machinery. Their inclusion sharply increased estimates of future electricity needs.

Kim Sung-hwan said confirmed semiconductor and AI projects alone required an additional 30 gigawatts, with the increase potentially exceeding 50 gigawatts by 2040 when electric vehicles and building heating were included. Separately, the 12th plan's general committee estimated 36.5 gigawatts of additional demand associated with the megaprojects by 2040, including 4.6 gigawatts for semiconductors and 11.9 gigawatts for AI data centers.

Lee Jin-young, a professor of electrical engineering at Pukyong National University, described the revision in annual consumption terms: the 2040 estimate rose from about 658 terawatt hours in the preliminary outlook to about 847 terawatt hours four months later. The increase of 189 terawatt hours is approximately 29% of the earlier figure.

The speed of that change has strengthened calls for a scenario in which demand grows more slowly. Possible reasons include weaker industrial investment, more efficient AI computing and delays in planned factories or data centers.

The ministry's official outlook records that participants in a public discussion requested a lower scenario reflecting industrial weakness and failure to achieve the 2035 emissions target. The ministry responded that information for such a scenario was insufficient and that too many scenarios could complicate decisions. The disagreement concerns how wide a range of possible futures the plan should prepare for.

How firm are the data center commitments?

The Energy Transition Forum says 7.4 gigawatts of an initial 8.4 gigawatts of data center demand, about 88%, comes from facilities intended for rental whose eventual tenants have not been determined. Occupancy matters because a proposed building's electricity requirement depends on whether customers actually install and operate computing equipment there.

Kim Jeong-ho, the Democratic Party lawmaker chairing the National Assembly's Climate, Energy, Environment and Labor Committee, has called for stronger evidence before uncertain projects enter the national plan. In an October 7 interview, he argued that some demand associated with global technology companies lacked investment plans or memorandums of understanding.

Kim explained his concern about building public infrastructure ahead of private commitments:

If we build generation facilities or power grids for foreign big tech firms that may or may not come, and they do not come, those become idle assets. It is right to exclude uncertain demand.

Semiconductor investments also face questions about timing. Companies must balance spending in South Korea against investment in the United States, while infrastructure preparations for proposed facilities in the southwest remain incomplete. Those decisions affect when factories begin drawing electricity.

Seok Gwang-hoon compares the current debate with the 2017 deliberations over Shin Kori reactors 5 and 6, now known as Saeul units 3 and 4. Construction already underway then created pressure on participants, he argues. Today, a forecast of industrial electricity shortages could exert similar pressure, turning discussion toward public acceptance of reactors rather than an examination of their necessity.

Power lines carry a large financial risk

Korea Electric Power Corporation plans to invest 83 trillion won in electricity networks through 2038: 72.8 trillion won for transmission and substations, and 10.2 trillion won for distribution networks. Transmission moves large amounts of electricity over long distances; distribution carries it onward to homes and businesses.

KEPCO data submitted to People Power Party lawmaker Na Kyung-won showed that 20 of 54 major transmission and substation projects in the 11th electricity plan were experiencing difficulties because of local opposition and permitting delays. That is about 37% of the listed projects.

Kim Sung-hwan told the October 8 parliamentary audit that the government would reexamine the grid plan during preparation of the 12th plan and look for spending that was unnecessary or could wait. The government is also considering greater use of existing lines, supported by electricity storage and limits on generation when the network is congested.

Consumers have a direct interest in the outcome. The minister said grid costs would be paid upfront and recovered over time through electricity charges and other means. If industrial consumption arrives later or is smaller than forecast, expensive infrastructure could operate below its intended use while its financing costs continue.

Local conditions also matter. A national surplus can coexist with difficulty delivering electricity to a particular industrial cluster. The official outlook identifies regional demand forecasting, linked to semiconductor and data center locations and grid planning, as a next step.

Could prices and efficiency reduce the need?

Rhie Sang-min, a senior research fellow at the Fiscal Reform Institute, argues that South Korea has surplus electricity for about 360 days a year outside the most demanding summer and winter periods. His concern is that building plants for brief peaks can leave consumers paying for facilities that spend much of the year underused.

President Lee Jae Myung supported electricity prices that change with supply conditions. At a national fiscal strategy meeting, he described the approach:

We need to change to a system in which rates are lower in times of surplus and raised in times of shortage.

Lee also said any increases needed to be accompanied by expanded energy vouchers for households with low incomes. Such support would help people who have little ability to move essential heating, cooling or other consumption to cheaper hours.

The official demand outlook already assumes that electric vehicles, heat pumps and industrial pricing by time of use will move 1.3 gigawatts away from summer peak periods. Moving consumption changes when electricity is needed, while efficiency reduces how much is needed.

Nuclear planning also sits alongside a target of at least 100 gigawatts of renewable capacity by 2030 and the retirement of coal generation. Kim Sung-hwan warned that AI growth and the megaprojects could threaten the 2035 greenhouse gas reduction target if clean electricity expansion and coal retirement fail to keep pace.

Earlier forecasts offer a warning about precision

Lee Jin-young's examination of the first 11 electricity plans found an average forecast error of 11.4% when looking 10 years ahead, compared with 1.1% for forecasts one year ahead. Errors went in both directions. The first plan projected 363 terawatt hours for 2011, when actual consumption reached 455; the sixth projected 579 terawatt hours for 2019, when actual consumption was 520.

Lee also identified a change in the relationship between economic growth and electricity use. Between 2001 and 2011, electricity consumption rose 1.3% for each 1% increase in economic output. Between 2011 and 2019, it rose only 0.6% for the same increase in output.

Demand management creates another uncertainty. Lee noted that the fifth through 11th plans subtracted savings targets equivalent to 14% to 16% of demand in their final years. He described a calculation for the 12th plan that subtracts 139 terawatt hours from 986 terawatt hours to reach 847. These figures combine a consumption forecast with an expectation that policies will successfully reduce it.

Construction schedules make mistakes costly to correct. Lee cited about 167 months, nearly 14 years, for a new large reactor and 13 years for grid infrastructure. His proposed response is to separate decisions that must be made early from those that can wait for stronger evidence.

International examples illustrate that approach. Britain procures electricity capacity through auctions four years and one year before delivery, with February 2025 targets of 43.7 and 7.5 gigawatts respectively. Lee also described how the US grid operator PJM screens near term data center demand for supply contracts or construction commitments. That screening, together with revisions to electric vehicle and economic assumptions, helped reduce its 2028 peak forecast by 4.4 gigawatts.

A timetable that has moved into 2027

The Electric Utility Act requires a new electricity plan every two years, looking 15 years ahead. The 12th plan covers 2026 to 2040. The 11th, finalized in February 2025 and extending through 2038, already included two large reactors and one small modular reactor, a smaller reactor design intended for modular construction.

The ministry's April forum announcement shows that demand discussions preceded the nuclear panel. It said the preliminary outlook followed four months of work and nine expert discussions. The public process has since developed through several stages:

• December 30, 2025, and January 7, 2026: The first two public forums examined the energy mix and the use of renewables alongside nuclear power.
• April 22, 2026: A third forum was scheduled for 10 a.m. at Kobaco Hall in Seoul's Yangcheon District, with attendance in person and an online livestream.
• August 2026: Kim Sung-hwan announced plans to begin nuclear deliberations and sought completion of the electricity plan before December.
• October 2, 2026: The nuclear public deliberation committee was launched.
• October 14, 2026: The committee agreed on the main agenda and exclusions.

The latest timetable is later than the minister's August goal. Kim Jeong-ho said the three months needed for deliberation, followed by government planning, put completion in the first half of 2027. An exact finalization date has not been announced.

Public opinion will be one input. A June survey of 2,000 adults, commissioned by the Korea Energy Information Culture Agency and conducted by Korea Research International, found that 82.9% supported including new nuclear plants in the supply plan and 79.2% considered new construction necessary. It took place before the megaproject announcements. The citizen panel will examine the issue through briefings and discussion, with its findings combined with expert assessments of supply and the grid.

Key Points

• The October 14 agenda focuses on nuclear power's role and share in meeting future electricity demand.
• The revised 2040 peak demand projection is 158.4 to 165 gigawatts, up by as much as 26.8 gigawatts from April.
• Critics want lower demand scenarios and stronger evidence that proposed data centers will attract customers.
• KEPCO plans 83 trillion won in network investment through 2038, while 20 of 54 major transmission and substation projects face delays.
• Efficiency, electricity pricing and shifting consumption away from peak hours are also part of the planning debate.
• The parliamentary committee chair expects the 12th electricity plan in the first half of 2027, following public deliberation.