Seoul's new industrial policy, branded 3S+1F (Strongholds, Speedily, Spearhead, Full), pledges roughly $900 billion for AI chips, datacenters, and physical AI. Its 18.4 gigawatt electricity demand by 2035 is the constraint without a roadmap.
South Korea just promised roughly $900 billion of chip, AI datacenter, and physical-AI investment over the next several years, and almost nobody noticed what the plan is actually asking the country's electric grid to do.
The plan, unveiled Monday at Cheong Wa Dae by President Lee Jae-myung, is branded 3S+1F. The acronym stands for "Strongholds" (regional fab clusters), "Speedily" (compressed timelines), "Spearhead" (new semiconductor forms), and "Full" (whole-of-government backing). The four pillars add up to roughly ₩1 quadrillion, or about $900 billion at current rates, in pledges from Samsung, SK hynix, SK Group, Naver, GS, and the state (The Register).
The largest chunk, roughly ₩900 trillion for semiconductors, has a credible delivery path: process roadmaps through 2 nm gate-all-around, high-bandwidth memory packaging treated as a known engineering problem, and timelines the country's two memory leaders are pulling inward. Samsung Yongin pulled forward 7 years, SK hynix Yongin pulled forward 12 (Korea Herald). The next-largest pillar, ₩550 trillion for AI datacenters from SK, GS, and Naver, is the one that quietly rewrites Korea's electrification math. Phase one alone is 8.4 gigawatts of compute. The full 2035 target is 18.4 gigawatts, a figure Science Minister Bae Kyung-hoon has priced into a ₩1,000 trillion (about $648 billion) total buildout (TechCrunch; SeDaily News).
For context, 18.4 gigawatts of continuous datacenter load is roughly equivalent to adding another South Korea's worth of industrial electricity demand stacked on top of the existing grid. Building it requires something the chip side of the plan does not: a generation and transmission roadmap that does not yet exist.
The Honam siting decision is best read as a power question. The four Honam memory fabs, Samsung and SK hynix collectively responsible for ₩800 trillion (about $518 billion) of the semiconductor pillar, drew the loudest political noise, with the opposition arguing the cluster in Gwangju and South Jeolla was chosen to reward a Lee-leaning region where an overwhelming majority of voters backed him in 2025. Cheong Wa Dae's own framing cited "untapped power resources" in Honam as the primary rationale, and Yongin and Pyeongtaek, the two existing mega-sites, are at electrical capacity. The political critique is fair shorthand. It also understates the engineering logic (Al Jazeera; Korea Herald).
The harder part of the constraint is what neither side of the political argument is talking about. Phase one, the 8.4 GW slice, is the easier 60 percent. Korean nuclear is restarting after the 2017 phase-out, combined-cycle gas can be permitted faster than most OECD peers, and existing baseload can be reallocated. The remaining 40 percent, the delta to 18.4 GW by 2035, is the clean-power problem. Korea does not have spare onshore wind geography, its offshore wind resource remains under-deployed, and its solar pipeline is congested on transmission. None of the 3S+1F pledges announced Monday include a generation-side roadmap, and the energy ministry has not yet aligned with the trade ministry on whether the additional baseload will come from gas, nuclear restart, or import contracts (SeDaily News).
The third pillar, covering physical AI and robotics, adds another layer of electricity demand on top of the datacenter loads. Physical AI and humanoid robotics manufacturing are both electricity-intensive. The pillar also has the fewest concrete dollar figures attached in available reporting, and the most indirect coupling to the rest of the plan.
What does have a clear delivery path is governance. Industry Minister Kim Jung-kwan has been designated a "chip czar," with the Ministry of Trade, Industry and Energy opening a ₩2 trillion semiconductor special account next year. The long-stalled K-Chips Act, Korea's analog to the U.S. CHIPS Act, cleared the National Assembly in a recent session. A separate IMEC-style advanced-packaging R&D hub has been ordered to Yongin, with the czar coordinating across ministries (SeDaily News).
Those institutional moves are real. The credibility test for the 3S+1F plan, however, is whether the Korean grid, supplemented by whatever new generation, transmission, and imported baseload the administration can assemble over the next decade, can deliver clean power at the scale the chip announcements implicitly assume. If Seoul can prove out the power curve, the chip plan executes, and Korea makes a credible run at the shift SK Group Chair Chey Tae-won described at the same event, from AI consumer to AI exporter. If the grid curve lags, Korea's most consequential industrial policy in a generation will be remembered as the one whose acronyms outran its transformers.