As the AI boom intensifies, fears over securing electricity grow deeper. Data centers are the factories of the AI era—and hippos that swallow electricity. According to the International Energy Agency (IEA), global data center power demand rose 17% in 2025, and power consumption by high-density AI data centers is surging. On top of this, Korea faces the practical challenge of operating semiconductor fabs and AI data centers simultaneously.
The production capacity of Samsung Electronics and SK hynix is directly tied to national competitiveness, and growing the AI industry also requires large-scale data centers. The war in the Middle East is accelerating this crisis. The World Bank projected that energy prices in 2026 will rise 24% due to the Middle East shock, reaching the highest level since 2022.
Expanding electricity supply is now an unavoidable option for maintaining national competitiveness. Nuclear power is needed, and renewable energy is needed. Transmission networks and storage facilities must also be expanded. But power plants cannot be built overnight. Transmission networks are even harder. Regional conflicts, community acceptance, environmental regulations, and permitting issues are all intertwined. That is why energy policy needs another pillar: demand management.
Korea's energy policy is managed in an integrated manner through the Basic Plan for Electricity Supply and Demand. The problem is that demand management policies are not receiving attention—not just from the public, but also from political circles and the government. At a time when securing energy is a matter of national competitiveness, policies on demand management must move beyond production-centered discussions and take their place as a major policy agenda. Making the electricity already in use less wasteful is as important as producing more. Energy policy in the AI era cannot rely on supply expansion and price adjustments alone. Demand must be managed with precision.
When a power crisis hits, the government launches massive energy-saving campaigns. Turn off the lights. Unplug the cords. Raise the air conditioning temperature. These are necessary, but not enough. The same goes for pricing policy. Raising electricity rates can reduce consumption, but it also pushes up consumer prices and industrial costs. It is even harsher on vulnerable groups and small and medium-sized enterprises.
Energy demand management cannot be left to campaigns alone. Households, small business owners, factories, and data centers each use electricity differently. The same conservation message cannot be sent to all. For households, neighbor comparisons and bill forecasts can be effective. For small business owners, guidance on reducing peak-time rate burdens is important. For factories, power usage by equipment must be shown alongside productivity data. For data centers, disclosure of power efficiency and incentives for demand response participation are needed.
This is where behavioral economics comes in. People do not act on rational calculation alone. They are more sensitive to immediate losses than to long-term national interests, and they respond more quickly to comparisons with others than to absolute numbers. A loss-aversion message like "You are using more electricity than similar neighbors" is more powerful than "Please save electricity." Conservation is not a matter of willpower but of choice architecture.
The U.S. case of Opower demonstrates this. Opower provides each household with information on its energy usage along with its electricity bill. According to a study of 600,000 electricity-consuming households, consumers reduced their power consumption by an average of 2% through comparisons with other consumers. By repeatedly making each consumer aware of their power usage, the company helped shape their electricity-use habits, and the resulting energy savings were sustained. What deserves attention is that people moved simply through this approach—a nudge (meaning "to poke gently with an elbow," referring to interventions that softly guide others' choices instead of using coercion or incentives) applied to energy consumption behavior. All it did was inform people of "your position (power consumption)" and suggest the "next action (saving electricity)."
Korea has sufficient foundations as well. Smart meters have already been deployed in more than 12 million households, and Korea Electric Power Corporation (KEPCO) provides real-time usage, estimated bills, excessive-use alerts, and neighbor comparison features through its PowerPlanner service. The problem is not a lack of infrastructure but a lack of utilization. PowerPlanner's subscription rate falls short of 10% of actual smart meter users. The platform is in place, but the link to behavioral change and savings outcomes is weak. Consumers must be actively informed so that they can specifically verify their energy savings rates, cost reductions, and peak power cuts. Furthermore, PowerPlanner must evolve beyond a simple power inquiry service into a policy platform that designs energy demand behavior in homes and industrial sites. A foundation must be built so that infrastructure leads to substantive energy savings.
Businesses must also view this issue as a nonmarket strategy. It is a strategic issue intertwined with electricity rates, supply chains, production capacity, data center locations, consumer burdens, and carbon policy. When companies build energy-saving models first, they not only reduce costs but can also drive changes in policy direction. Smart building demand management, factory peak-power reduction, data center power efficiency disclosure, and energy-saving support for partner firms are excellent policy agendas companies can propose.
If AI triggered the power crisis, the solution can also be found in AI. The idea is to use AI as the "brain of savings" that reads consumption patterns and cuts out the points of waste. There is no more practical alternative than AI for managing real-time power demand through individualized targeting. It should propose optimal conservation habits for households, identify wasted electricity in factories, and apply load-balancing algorithms to data centers.
Expanding supply is like enlarging the water tank. But enlarging the tank alone will not solve the crisis. We must first examine where the water is leaking. Today's holes are the absence of information and the lack of compensation. The government must now devise AI-driven and behavioral-economics-based policies to make people use less electricity—just as much as it works to produce more.
