# South Korea is treating nuclear construction speed as an AI infrastructure problem and it's right > Source: > Published: 2026-06-30 06:01:33+00:00 *Seoul's announcement that it will try to cut nuclear plant build times from a decade to something shorter isn't energy policy , it's AI industrial strategy, and other governments are watching.* On June 30, presidential chief of staff Kang Hoon-sik told reporters that South Korea will examine ways to shorten nuclear plant construction timelines , currently running nine to ten years , to meet the power demands of its fast-expanding AI and semiconductor industries. Details will be folded into the government's upcoming long-term electricity supply plan. That single sentence reframes a century-old infrastructure problem. Nuclear buildout speed is now, formally, an AI competitiveness issue. The timing is not coincidental. A day earlier, President Lee Jae-myung presided over what his government called a national "great leap" event, unveiling more than 1,000 trillion won , roughly $651 billion , in commitments across semiconductors, AI, and data centers. Samsung Electronics and SK Hynix will anchor two new chip fabrication sites in the country's southwest. A separate quadrillion-won data center investment was announced alongside it. That's a lot of electricity to promise and not yet have. South Korea's 11th Basic Plan for Electricity Supply and Demand, finalized in February 2025 and running through 2038, already targets nuclear's share of the grid rising from roughly 25 gigawatts today to 35 gigawatts by the end of the period , around 35.6% of total generation. The plan calls for two new large-scale reactors with combined capacity of 2.8 gigawatts, slated for completion in 2037 and 2038, plus a 0.7-gigawatt small modular reactor by 2035. An SMR Special Act, stalled in parliament for two years, finally passed in February 2026, signaling cross-party consensus rather than a one-administration bet. The problem is the timeline. Two new reactors finishing in 2037 and 2038 do almost nothing for data centers coming online between now and 2030. That gap is what Kang's announcement is actually about. When a government says it wants to "examine ways to shorten" a nine-to-ten year build time, it's acknowledging that the existing permitting and construction pipeline was designed for a power-demand curve that AI has already made obsolete. Frankly, South Korea is in a stronger position than most to act on this. It has operational experience with nuclear at scale , 24.7 gigawatts of nuclear capacity already running , a domestic reactor builder in KEPCO, and a public that has shifted toward support for nuclear as AI power demand has become visible and real. The February SMR act didn't pass because politicians suddenly got brave. It passed because the demand argument became undeniable. ## Whether Other Governments Follow The pressure is not unique to Seoul. Global data center electricity consumption is projected to exceed 1,000 terawatt-hours by the end of 2026, according to industry estimates , roughly equivalent to Japan's total annual electricity use. In the United States, total data center energy demand is estimated to nearly double from 80 gigawatts in 2025 to 150 gigawatts by 2028. Arizona has moved legislation to accelerate permitting for small modular reactors co-located with data centers. The UK and US have also agreed to cooperate on accelerating nuclear development. France and Poland are expanding nuclear programs of their own. What South Korea has done that others haven't, at least not yet, is make the connection explicit at the head-of-government level. Kang's statement didn't frame nuclear acceleration as a climate or energy security measure. It framed it as a response to AI. That's a different kind of political commitment, because it ties nuclear delivery schedules directly to a concrete industrial outcome , chip exports, data center revenue, sovereign AI capacity , rather than to an abstract emissions target twenty years away. You can hold a government accountable to a chip factory's power-on date in a way you can't to a 2050 net-zero pledge. The harder question is what "shorten" actually means in practice. Cutting a nuclear build from ten years to seven requires regulatory reform, pre-approved site designs, faster environmental review, and workforce that doesn't bottleneck on licensed welders. South Korea has advantages in all of these , KEPCO's APR1400 design is already proven and licensed , but the government hasn't yet specified whether it's targeting three years off the schedule or one. The details will come in the electricity supply plan. Until then, the announcement is a signal more than a blueprint. What it signals, though, matters. AI's power appetite has now forced a sovereign government to treat decade-long nuclear construction timelines as an engineering problem to be optimized, not a political reality to be managed. That's a shift in how energy infrastructure gets framed at the policy level, and it's one other governments with serious AI ambitions will have a hard time ignoring. **Also read:** [Vibe coding has rewritten the rules on who gets to build a startup and billions in VC money are betting it sticks](https://startupfortune.com/vibe-coding-has-rewritten-the-rules-on-who-gets-to-build-a-startup-and-billions-in-vc-money-are-betting-it-sticks/) • [The Magnificent Seven just lost $2 trillion and the market is asking a question big tech cannot yet answer](https://startupfortune.com/the-magnificent-seven-just-lost-2-trillion-and-the-market-is-asking-a-question-big-tech-cannot-yet-answer/) • [The AI infrastructure boom is turning Q2 2026 into the market's best quarter in six years and founders should pay attention](https://startupfortune.com/the-ai-infrastructure-boom-is-turning-q2-2026-into-the-markets-best-quarter-in-six-years-and-founders-should-pay-attention/)