AI's potential to transform the renewable energy market
“An increase in AI data centers does not automatically lead to an increase in renewable energy.”
Kim Yu-chang, head of a division at KEI Consulting, emphasized that although the expansion of AI data centers is driving a sharp increase in electricity demand, it does not directly translate into growth in the renewable energy market. Kim said the issue of supplying data centers with reliable electricity must be distinguished from procuring renewable energy to meet RE100 commitments.
Kim emphasized that physical infrastructure, including the power grid and energy storage systems (ESS), must support reliable data center operations. Regarding the government's Korean Green Transformation (K-GX) strategy and its target of 100 GW of renewable energy by 2030, Kim also pointed to the need for an institutional framework that encourages private investment.
“Securing data center power and RE100 procurement are separate issues; LNG and ESS are both needed”
The most important issue in an AI data center business is securing electricity on time. Buildings and servers can be prepared in one to two years, but a data center cannot operate if its grid connection is delayed.
Securing grid access is also identified as a major challenge in South Korea. According to Kim, transmission line construction is delayed by an average of about seven years, and only four of 195 applications for electricity use by data centers in the Seoul metropolitan area passed between August 2024 and June 2025.
Overseas, a model known as BYOP, or Bring Your Own Power, has emerged to address such grid connection delays by securing on-site generation facilities. Under this model, data centers begin operating with their own generation before the grid connection is completed, then use grid electricity alongside on-site generation afterward. Renewable energy procurement, by contrast, is different in nature from electricity supply.
Even if a data center obtains recognized renewable energy use through a remote power purchase agreement (PPA) or renewable energy certificates (RECs), the electricity it actually consumes is grid power comprising a mix of nuclear, LNG, coal and renewable generation. Kim said, “Physical supply is a question of whether electricity can reach a data center on time, while renewable energy procurement is a contractual question determining what kind of electricity its consumption is recognized as.”
Kim explained that long-term contracts linked to new power plants are important if renewable energy procurement is to lead to actual investment in generation facilities. Rather than simply buying certificates from existing facilities, expanding PPAs lasting 10 to 20 years would allow developers to raise financing and build new power plants. Reliable electricity supply for data centers is a separate issue that also needs to be addressed. Solar and wind output varies with the weather, and generation gaps can last several days. Conventional short-duration ESS alone cannot readily cover such gaps.
Kim proposed combining LNG generation and ESS to provide reliable electricity for data centers. LNG generation would provide firm output, inertia and voltage support, and sustained supply, while ESS would respond to sudden load fluctuations and momentary outages.
“What data centers require is electricity that is uninterrupted and stable,” Kim said. Kim explained that a practical approach would be to secure near-term supply reliability through LNG generation and ESS while meeting long-term carbon reduction targets by gradually increasing the share of long-term renewable energy PPAs.
Kim took a cautious position on mandatory renewable energy procurement for data centers, arguing that increasing the volume available in the market so companies can procure it voluntarily should take priority over mandating purchases. Instead, Kim identified data centers' power usage effectiveness (PUE) as an area requiring regulation. Since building layouts, cooling methods and electrical equipment are largely determined during construction, Kim explained, applying efficiency standards from the initial design stage would be effective.
Kim considered it inefficient to impose purchasing obligations on developers because renewable energy procurement is often handled by operators and their tenant customers. Kim proposed requiring operators to disclose renewable energy procurement targets and performance, as well as transparently disclosing ancillary costs such as PPA network charges. Kim also identified a need to clarify how recognized renewable energy use should be allocated between operators and tenant customers.
“Private investment and market reforms are key to 100 GW of renewable energy”
Kim emphasized that achieving the government's K-GX strategy and its target of 100 GW of renewable energy by 2030 would require moving beyond publicly led deployment policies and expanding private capital participation. “Looking at K-GX and the first Basic Plan for Renewable Energy, the 100 GW target has a publicly led structure,” Kim said. “The mechanisms to attract private investment appear considerably inadequate.”
Kim drew attention to a structure in which the government determines generation sites, buyers and prices, while public institutions provide demand. Kim said support for expanding private companies' long-term PPAs amounts to little more than extending network charge assistance. The scale of capacity needed to meet the deployment target is also substantial. Kim argued that increasing renewable energy capacity from 37 GW in 2025 to 100 GW in 2030 would require an annual average of about 12.6 GW of new capacity. Considering that recent annual additions have been around 3 to 4 GW, this would mean expanding at three to four times the current pace.
Kim explained that policy predictability must be secured so private developers can make long-term investment decisions. In particular, Kim said clear timetables and rules are needed for changes affecting project profitability, including the abolition of the Renewable Portfolio Standard (RPS), consolidation into a single competitive auction framework and regionally differentiated electricity tariffs.
“For private investors making a 20-year investment decision, the greatest risk is not low prices but changing rules,” Kim emphasized. Kim also saw a need to change how auctions are run to reduce generation costs. Rather than using an auction price ceiling to push prices down, Kim argued, the government should allow sufficient competition to establish appropriate prices reflecting development and financing costs.
Regarding the changes AI could bring to the electricity market, Kim acknowledged the technology's potential while emphasizing the importance of market rules. “AI will be used to improve operational efficiency to a very high degree,” Kim said. “But if South Korea's rules remain unchanged, efficiency will improve, but it will be difficult for AI to fundamentally transform the market.”
AI optimization and electricity market incentives
Generation forecasting, ESS charging and discharging control, and virtual power plant (VPP) operations were already using optimization technologies before AI emerged. Recent AI differs in its ability to learn from more data simultaneously, improving prediction accuracy, and to integrate and control numerous distributed resources in real time.
However, the economic value created by these technologies depends on electricity market price signals. Kim explained that AI optimization can deliver greater economic benefits when there is an opportunity to profit by buying electricity during low-price periods and using it during high-price periods.
A prominent example is Kraken, the energy management platform of the UK's Octopus Energy. Kraken automatically adjusts electric vehicle charging and electricity use in response to electricity price changes. The UK has large fluctuations in wholesale prices, and those changes are reflected in consumer tariffs. South Korea, by contrast, has smaller wholesale price fluctuations than the UK, and retail tariffs are separated from wholesale prices. This makes it difficult to expect the same economic benefits as in the UK, even if the same AI technology is adopted.
This does not mean that adopting a UK-style pricing system is the answer. Risks arising from electricity price fluctuations could be passed on to consumers.
“If improving energy market efficiency means sacrificing the welfare of industry and ordinary consumers, there is no particular reason to change the system,” Kim said. Kim's position is that the results of introducing AI should ultimately be verified against existing optimization methods under the same market rules. Kim added that generation forecasting errors, settlement costs, ESS operating revenue, reductions in battery degradation and renewable energy curtailment should be compared using the same periods and the same resources.
Original Korean article: Korea Economic Daily
English brief by Korea Wind Intelligence, koreawindintel.com
