South Korea just went big on AI infrastructure, unveiling a “3S+1F” plan backed by over ₩1 quadrillion in AI-related investment, roughly $900 billion, according to a government and industry announcement reported by The Register. The headline number is not window dressing. It is tied to a specific bottleneck: datacenters with combined capacity of 18.4GW that are supposed to be built by 2035.

That power plan matters because compute is the real fuel for what comes next. Under “3S+1F,” Seoul says it will “Speedily build” fabs in regional hubs called “Strongholds,” and also “Spearhead” new forms of semiconductors needed by growing markets. It is also pushing robotics development “ASAP,” with the explicit ambition of rivaling China as the world’s unchallenged center for robots, alongside plans to build “Physical AI models,” described as doing for movement what LLMs do for text.

If you are an executive watching the AI supply chain, the interesting part is how the strategy ties together manufacturing scale, compute capacity, and product direction. Semiconductors, datacenters, robotics, and AI models are often treated as separate business threads. Here, the government is trying to stitch them into one industrial push, anchored by semicon fabs and powered by grid-scale electricity. The Register describes the datacenter buildup as enabled by massive investments in clean energy generation and a grid designed to move electrons efficiently, “literally and metaphorically electrifying the South Korean economy.”

The plan’s industrial choreography also signals where Seoul expects leverage. SK Hynix and Samsung are both named as part of the effort, each committed to spending vast sums on new fabs. LG and Hyundai will also invest. In other words, this is not a government-led capex story that relies on promises from distant vendors. It is a coordinated push involving the firms that already sit at the center of memory and chip manufacturing capacity.

There is also a governance and incentives angle worth noting. The Register says President Lee Jae-myung pitched the plan as both social and economic policy, including the idea that it will build the industrial base needed for AI faster than any other country and spread wealth beyond Seoul. That framing matters for boards and investors because it turns an industrial policy into a political mandate. When capex is treated as national competitiveness and social distribution, it can become harder to pause, harder to reroute, and easier to defend as a long-cycle project even when market conditions fluctuate.

Strategically, the plan is positioned as partly an answer to China. The Register calls out that South Korea’s announcement uses the rather Chinese phrase “great leap forward” several times. That word choice is ironic on its face, because the scheme is intended to ensure South Korea can compete with China, but to do it with local factories rather than South Korean tech giants investing in the Middle Kingdom. Put simply, the government wants chips made at home for the AI buildout, not just capital deployed where growth or demand might already be booming elsewhere.

The second-order implication for executives is that this can reshape competition around the entire AI stack. If “Strongholds” for fabs accelerate, and datacenters scale to 18.4GW by 2035, the bottleneck is not just model training. It becomes domestic throughput: manufacturing capacity, power availability, and the ability to turn compute into robotics and “Physical AI models.” That, in turn, affects procurement strategies, partnership choices, and where companies look for risk. When a country coordinates power, chips, and model direction, the advantage can compound for years.

For peer regions and companies making capital allocation decisions, “3S+1F” is a reminder that AI is not only an algorithms story. It is an industrial base story, and Seoul is trying to write the operating manual. The plan aims to emerge as a “K-Semiconductor powerhouse” and a global leader in robotics and AI, with government and industry support underpinning the whole package. If it works as intended, it could shift the center of gravity for AI hardware and robotics manufacturing, and force rivals to decide whether to match scale, or accept a slower path to industrial relevance.