South Korea drafts the AI petrostate

BURIED in Kim Yong-beom's 2,500-word Facebook post late on Monday night was a phrase that the next morning would wipe more than $300bn from the Kospi in a single session of trading. The phrase was "national dividend," and Mr Kim, South Korea's presidential chief of staff for policy, is not a man given to populist flourishes. He is a veteran economic bureaucrat, a former vice finance minister, and a principal designer of President Lee Jae-myung's domestic agenda. His post argued, at length and with care, that the windfall accruing to Samsung Electronics and SK Hynix from the global AI infrastructure build-out should be "structurally returned" to the Korean public — and that the principle, debated and refined first in Seoul, "could eventually become a standard for nations in the AI age."

The market did not parse it that calmly. The Kospi, which had opened Tuesday at a record 7,999, fell as much as 5.1% within hours. Foreign investors pulled 5.6tn won from the bourse in a single session. Samsung and SK Hynix shares dropped, and within the day Mr Kim had clarified that he was referring to "excess tax revenue" rather than a new levy on corporate profits. Cheong Wa Dae, the presidential office, called the post a "personal opinion." The opposition People Power Party called for Mr Kim's dismissal. Mr Lee, by contrast, dismissed the press coverage as "fake news" — which is to say, he did not walk it back.

If Samsung and SK Hynix hit their 2026 forecasts (330tn won and 239tn won in operating profit, respectively), their combined corporate tax bill alone could exceed 100tn won, surpassing the entire national corporate tax collection Korea's finance ministry has projected for this year. Samsung, by Bloomberg's estimate, will rank second only to Nvidia among the world's most profitable companies, ahead of Apple and Alphabet. By Nomura's projection, the two firms' combined operating profits could reach 600tn won this year — roughly a quarter of South Korean GDP. By the standards of every previous semiconductor cycle, this is an unfamiliar order of magnitude. By the standards of Norway in the 1970s or Finland in the 1990s, it is the beginning of something the country has not yet built the institutional architecture to absorb.

Seoul searching

Three things are on the table at once, and the Lee administration's working position is somewhere underneath them. The first is the "national dividend" itself, which is less a program than a principle: that AI-era rents accrue partly to infrastructure built over half a century with public money, and so should be partly returned. The form is undetermined. Mr Kim has floated basic income for rural communities, expanded pensions, support for young entrepreneurs, and education spending for the AI transition. What survives the legislative process is what foreign capitals will be watching.

The second is the labor question, which is moving on its own clock. Samsung's union has set May 21st as the start of an 18-day strike unless the company commits to higher wages and bonuses tied to AI profits — 15% of operating profit, in the union's framing, is the right share for the 30,000 chip-division employees whose work produced the windfall. SK Hynix already routes 10% of operating profit to a performance bonus pool. Analysts reckon an 18-day stoppage could cost Samsung between $6.9bn and $11.7bn in lost foundry output. The state retains an emergency arbitration option under Article 76 of Korea's labor law, used only four times since 1969. The Lee government has not yet reached for it.

The third question (and the one that gives the first two their international consequence) is whether South Korea now functions as a memory petrostate. Two firms that together approximate a quarter of national output is the kind of concentration that has only one historical analogue in the OECD: Nokia at its 2000 peak, when the firm accounted for roughly 70% of Finland's market capitalization and 230% of Finnish GDP. Within a decade, Nokia had unwound. Taiwan is now running an analogous experiment with TSMC, whose market value sits at roughly twice Taiwanese GDP and whose share of the local index has climbed past 40%. The Lee administration is reading both cases. So, presumably, is everyone else.

The Norway parallel is the cleanest, and the one Mr Kim's "once-in-a-generation historical opportunity" rhetoric most plainly reaches for. Norway discovered oil in 1969; parliament established what would become the Government Pension Fund Global in 1990; the first deposits arrived in 1996; the 4% spending rule was codified in 2001. Three decades from discovery to a functioning fiscal stabilizer. The fund is now worth more than $2tn, roughly $390,000 per Norwegian. The lesson Norwegian officials still describe in interviews is that the country tried, in the 1970s, to spend the windfall directly: pension cuts, agricultural subsidies, industrial policy. The macroeconomic damage took the better part of a decade to repair; the fund came later, and only after the mistake.

Mr Kim is asking what happens if Korea does not wait for the mistake. The market reaction this week is the answer to a narrower question: can a government with a strong legislative majority, a redistributionist president, and a chief policy aide willing to write 2,500 words on Facebook design the institutional architecture before the cycle peaks? It is not yet clear. The chaebols, after four decades of every Korean government trying to discipline them, remain the dominant institutional actor in the political economy. The People Power Party will frame any structural redistribution as a "communist rationing economy" (its chair, Jang Dong-hyeok, has already done so). The market, as Tuesday demonstrated, is jittery about the difference between excess tax revenue and excess corporate profit even when the policy chief is careful to draw it.

Still, the question is now on the table in a way it is not yet on the table in Washington, Brussels, or Tokyo. Whatever architecture the Lee administration ends up with — a fund, a windfall tax, a UBI pilot, a hybrid — will set the rough template that Taiwan, Ireland, the Netherlands, and eventually the United States will examine when their own AI-era concentrations force the question. Mr Kim's post will not be the last 2,500 words written on the subject. It might, with some luck, be the first.