South Korea Says AI Could Add Power Demand Equal to Around 20 Nuclear Reactors
South Korea expects artificial intelligence, data centers and semiconductor expansion to add 25 to 30 gigawatts of electricity demand in the coming years. To put that into perspective, the country’s energy minister says that amount of power is roughly equivalent to the capacity of about 20 modern nuclear reactors.
Table of Contents
01 Event
South Korean Climate, Energy and Environment Minister Kim Sung-whan told Reuters that AI-driven projects from major chipmakers Samsung Electronics and SK Hynix, along with new data centers, will sharply increase national electricity requirements.
The government is preparing an updated long-term energy roadmap extending to 2040. One of the central questions is whether the country should build additional nuclear reactors to support rising demand.
02 What Changed?
Several large industrial projects have been accelerated. Samsung and SK Hynix are part of an 800 trillion won, or roughly $590 billion, semiconductor “mega project,” with facilities that had once been planned for around 2045 now expected much earlier in the 2030s.
At the same time, global technology companies are interested in building additional data centers in South Korea because of its reliable electricity system. That places new demand on a grid that must also support electric vehicles and the shift from gas heating to electric systems.
03 Why It Matters
AI is often discussed as software, but large-scale AI is fundamentally an industrial activity. Chips must be fabricated in power-intensive factories, data centers must run continuously, and cooling and networking equipment consume additional electricity.
That creates a planning problem. A country can approve a data center quickly, but new transmission lines and power plants may take years. If electricity supply does not grow alongside computing demand, prices rise, grid connections are delayed or projects move elsewhere.
04 What It Means for You
For consumers, higher industrial electricity demand can matter if infrastructure costs are spread across the wider power system. Governments and utilities must decide who pays for new generation and transmission: data-center operators, taxpayers, ratepayers or some combination.
The timing challenge is especially important. Chip factories and data centers can be built faster than major power plants or transmission corridors. If demand arrives first, grid operators may have to rely on existing gas or coal generation longer than planned, curtail new connections or invest in expensive short-term capacity.
That creates a tension between industrial policy and climate policy. South Korea wants to strengthen its semiconductor and AI position while also reducing reliance on coal. Achieving both goals requires enough new low-carbon electricity to replace retiring generation and serve new computing demand at the same time.
Energy security also becomes part of technology competitiveness. A country with reliable power can attract data centers and fabs even if its electricity is not the absolute cheapest. For AI companies, an outage can be more expensive than paying a modest premium for dependable supply, which is why grid reliability increasingly influences where large computing projects are located.
For technology users, the connection is indirect but real. Cheap AI depends partly on cheap, reliable electricity. If power becomes a bottleneck, the economics of AI services become less favorable no matter how efficient the underlying software becomes.
Earnyx has previously examined the wider cost of AI data-center expansion for power and water systems. South Korea provides a particularly clear national-scale example of that pressure.
05 Numbers + Context
The expected AI-related increase is 25–30 GW. South Korea currently operates 26 nuclear reactors, which provide just under one-third of its electricity. If the entire additional AI load were supplied by nuclear power alone, it would require capacity comparable to about 20 reactors.
Renewables currently supply about 11% of national power, while coal provides around 29% and gas around 27%. South Korea plans to phase out coal by 2040, which means new AI demand is arriving while the country is already trying to transform its generation mix.
A state-backed survey cited by Reuters found that 79% of 2,000 respondents supported including new nuclear plants in the country’s future energy roadmap, although local opposition remains significant in some potential host communities.
06 Earnyx Takeaway
The “20 nuclear reactors” comparison makes the AI infrastructure problem easy to understand: computing growth is becoming large enough to affect national energy planning.
That does not mean South Korea will build 20 reactors specifically for AI. It means AI-related electricity demand is now large enough that governments must make decisions normally associated with heavy industry. The next AI race will not be won by models and chips alone. It will also depend on who can build reliable power fast enough to keep them running.
