As artificial intelligence (AI) – the defining trend of our era – demands unprecedented volumes of electric power, major energy markets worldwide face the daunting task of resolving complex infrastructure hurdles within a tight timeframe. These critical tasks include optimizing the power generation mix, making substantial investments in transmission networks, modernizing aging electrical equipment, enhancing overall system stability and navigating the energy transition. Historically, structural shifts in power markets unfolded over extended horizons based on long-term planning. However, because the paramount requirement of the AI revolution is rapid, short-term energy supplies, this mandate has become far more burdensome for energy authorities and market participants. Korea is no exception.

On June 29, the government officially announced its strategic "Three Major Megaprojects" initiative, which prominently positions semiconductor manufacturing clusters and AI data centers at its core. According to published plans, operating the four proposed semiconductor plants in Gwangju – for which completion dates have not been finalized – will require an additional 6.3 gigawatts (GW) of generation capacity. Furthermore, powering AI data centers slated for construction in the region outside the Seoul Metropolitan Area (SMA) will demand an additional 18.4 GW of capacity by 2035. When combined with the 15 GW required to run the 10 semiconductor facilities currently under development in the Yongin cluster, total new power demand driven exclusively by the semiconductor and AI sectors approaches nearly 40 GW. This requirement represents approximately one-quarter of Korea’s total currently installed power capacity. Consequently, an uninterrupted, secure and resilient power supply serves as the non-negotiable prerequisite for the success of these megaprojects.

The government intends to mobilize all available energy resources, including renewable energy and nuclear power. Under the announced framework, electricity needed for the Gwangju cluster will be supplied by leveraging the Honam region’s abundant solar and wind resources alongside existing facilities like the Hanbit Nuclear Power Plant. In parallel, the government also plans to expedite the construction of transmission lines connecting power plants directly to industrial complexes while reviewing measures to shorten construction schedules for new nuclear units. For the Yongin cluster in the SMA, the strategy involves drawing upon power generated by liquified natural gas (LNG) plants. To facilitate this transfer, authorities plan to increase existing grid capacities while accelerating new grid construction through undergrounding methods to minimize delays. In the case of AI data centers, while the plan utilizes a diverse energy mix spanning renewables, nuclear power and LNG generation, the government emphasizes expanding distributed energy systems — ensuring electricity produced locally is consumed locally. Additionally, to address the intermittency of weather-dependent renewables, the plan incorporates the deployment of large-scale Energy Storage Systems (ESS).

Semiconductor plants represent an extraordinarily sensitive class of power consumers; even a momentary outage can result in the total destruction and disposal of an entire production batch of silicon wafers. Similarly, AI data centers are complex demand points to service due to continuous, sudden and extreme load fluctuations. Energy delivery to these facilities must satisfy rigorous criteria regarding not only the quantity of power supplied, but also the strict quality and stability of that power. Given these stringent technical parameters, a wide array of policy proposals is emerging across the energy sector. For example, a sharp debate has formed between advocates arguing for a substantial increase in renewable energy capacity to meet the nationally determined contributions, and opposing voices contending that the government must prioritize a stable energy supply, even if that necessitates revising climate targets. Furthermore, some experts suggest adopting a dual-track approach combining variable solar and wind power with grid power backed up by large-scale nuclear or LNG facilities. Others recommend utilizing all available energy assets including Honam nuclear power, large-scale ESS installations and LNG generation in a technology-neutral manner. Meanwhile, additional proposals advocate extending the operational lifespans of existing power units nearing or past their design lives, or constructing an East-West high-voltage transmission grid to expand capacity in the southwestern region and even integrate nuclear power generation from the Yeongnam region.

While most current proposals remain focused on how to supply this increasing demand for power, key stakeholders must recognize that any supply strategy lacks practical feasibility unless anchored in precise demand forecasting and paired with a clear mechanism for allocating rising costs. With respect to demand forecasting, power demand from semiconductor facilities and data centers — which fluctuates rapidly according to volatile AI market conditions — must be rigorously separated from traditional forecasting models based on historical macroeconomic growth data. Policymakers must carefully analyze the extent to which corporate investment plans and shifting policy objectives should be treated as confirmed, guaranteed demand. Moreover, establishing granular, year-by-year demand predictions in the absence of finalized implementation roadmaps represents a primary technical challenge. Without precise demand modeling, even the most elaborate supply plan remains essentially unusable.

With respect to cost allocation, rapidly expanding renewable energy inherently drives up system integration costs, including expenses for ESS installations to ensure grid stability and massive capital expenditures for new transmission infrastructure. Establishing an optimal supply mix balancing economic viability with environmental sustainability requires rigorous cost analyses for each source. Crucially, transparent policy discussions regarding how to distribute these additional cost burdens — particularly concerning their impact on consumer electricity tariffs — must proceed in tandem with technical planning.

The Three Major Megaprojects represent an unprecedented national enterprise valued at approximately 4,000 trillion won ($2.817 trillion), holding profound potential to build a foundation for Korea’s sustainable growth. To maximize the probability of success, realistic demand forecasting and rational cost allocation frameworks must be fully integrated into energy policy from the very first stage of design.

Kim Sung-woo, head of Environment & Energy Research Institute at Kim & Chang, is a board member of KETEP.