The scale of the AI boom is becoming easier to understand when measured in power plants instead of GPUs. Reuters reports that South Korea is nearing an agreement for more than $100 billion of U.S. investment, including construction of as many as eight nuclear reactors and a natural-gas plant in Texas, with much of the electricity intended for AI data centers.
The first reactor would reportedly use Westinghouse↗ technology, while later projects could use designs from Korea Electric Power Corp. Final details are still under negotiation, so the number and timing of plants can change. The investment sits inside a broader U.S.–South Korea economic arrangement involving manufacturing commitments and tariff treatment.
The strategic logic is bigger than electricity. AI infrastructure now binds together chips, memory, data centers, transmission capacity, cooling, natural gas and nuclear supply chains. South Korea is unusually well positioned because Samsung and SK Hynix are central to advanced-memory production, while Korean engineering companies have deep experience building nuclear plants.
This also changes the geography of technology investment. For years the scarce resources in software were talent and capital. For frontier AI, electrons, transformers, substations, land and construction schedules are becoming equally important. A data-center project can have money and GPUs and still wait years for grid capacity.
For entrepreneurs, the second-order opportunities may be more interesting than another model wrapper: energy optimization, cooling, grid software, construction logistics, data-center operations, compliance and tools that make expensive compute more productive.
South Korea is preparing a U.S. investment package worth more than $100 billion that could include as many as eight nuclear reactors plus a Texas natural-gas plant, according to a Wall Street Journal report cited by Reuters. The striking part is the purpose: much of the new power would support AI data centers.