The Ghost of the Five-Year Economic Development Plan
An industry built on connecting each other's strengths — seed money of 200 million won each for more than ten domestic fabless startups… AI [data centers] that prioritize the use of domestically made chips…
AI Summary
A proposed state-led AI semiconductor five-year plan in South Korea is drawing comparisons to the government-directed industrial policies of the 1960s–80s, with critics questioning whether its scale and logic fit today's global AI landscape. The plan's funding — just 200 million won per fabless startup — is widely seen as negligible against the capital demands of global chip and AI infrastructure competition, compounded by South Korea's already-lagging cloud competitiveness. The author argues that AI rewards ecosystem collaboration over self-sufficiency, and that politically appealing slogans like 'domestication' and 'self-reliance' are no substitute for genuine industrial competitiveness.
The idea of a state-led five-year plan for AI semiconductors is in the air. Items line up one after another: injecting 200 million won in seed money into each of more than ten domestic fabless startups, building AI-dedicated data centers that give priority to domestically made chips, and a 'K-Ontology Project' designed to let AI leverage sensitive data. Just reading the names, it all feels like something we've seen somewhere before.
The fundamental question stems from exactly that. Isn't this ultimately the specter of state-led industrial growth in the mold of the 'Five-Year Economic Development Plan'? The underlying idea is to cultivate AI semiconductors using the same playbook where the government designated sectors and dispersed funds to build industries — but neither the era nor the industry is what it was back then.
The limits of financial firepower are glaring from the outset. Two hundred million won per fabless company — when you consider the order of magnitude of global semiconductor and AI infrastructure competition, it is almost embarrassing to call it seed money at all. What stings even more is the reality that South Korea's cloud competitiveness is already significantly behind. The very foundation beneath the ambition to build chips is shaky.
The essence of the AI industry is not self-sufficiency. It is a game of connecting the distinct strengths that each country and each company possesses. In practice, cases where a third-party company took another's equipment and platforms to build a superior model have been read not as any single nation's solo victory, but as a victory for the entire ecosystem. Yet the Korean-style vision keeps drifting toward walling things off and trying to do everything from within.
The word 'domestication' is always politically sweet. It can bundle the justification of self-reliance, the packaging of national security, and the promise of domestic jobs all in a single sale. But sweet justifications and actual industrial competitiveness draw from different accounts. The specter of the five-year plan always looks plausible on paper. The problem is that the market does not read that paper.
Source: Naver Premium pickool ("Jensen Huang Has Left — The Real Face of South Korea's AI, as Seen by a Former Minister, a Lawmaker, and Two Founders")
This article was automatically translated from the Korean original by AI. For the authoritative version, read it in Korean.
한국어 원문 읽기 →