South Korea has announced an unprecedented plan. President Lee Jae Myung’s government says it will mobilise 1,800,000 billion won, roughly 1,170 billion euros of public and private investment, to build a new semiconductor manufacturing hub and a network of data centres for artificial intelligence. The figure is large enough to stop meaning anything, which is precisely why it works so well in a headline.
Start with what is actually inside it. The core is 800,000 billion won, around 520 billion euros, that Samsung Electronics and SK hynix will put into four new chip fabrication plants, two each, according to industry minister Kim Jung-kwan. They are destined for the south-west of the country, a region historically less industrialised than the Seoul metropolitan area, and the government is promising accelerated permitting to cut construction time and cost. Alongside that sits a programme worth 1,000,000 billion won, about 650 billion euros, for AI infrastructure by 2035, with a first phase of 550,000 billion won, roughly 360 billion euros, to be invested by 2029 in new data centres. That first phase sits inside the AI programme; it is not an extra line to add on top. And all of it is announced and still-to-be-mobilised money, public and private, not money already spent. A plan is a promise with a stamp on it.
Then there is the sentence that will travel further than any of the figures. “Only speed will allow us to survive,” Lee said, arguing that the country must secure the fundamental elements of artificial intelligence faster than anyone else. It is a good line. It is also where two very different kinds of speed get quietly folded into one word.
Because the speed the government is actually offering is administrative. Faster authorisations, shorter paperwork, fewer months lost in offices. Useful, genuinely. It is not the speed that builds anything. Analysts and academics warn that creating an industrial ecosystem away from the established clusters will take years, on top of heavy spending on infrastructure, water supply and qualified labour. Those are risks, not failures already on the record. They are also the part of the plan that never opens a press conference. A signature can shorten an authorisation; it cannot pour concrete, lay a water main or train an engineer in its place. Those run on their own clock, and a faster permit does not stand in for doing the work or training the people.
There is a political logic here too, and it deserves to be named rather than admired in the abstract. Moving the fabs to the south-west is part of Lee’s strategy to rebalance economic development away from the capital region. That is a legitimate goal and a difficult one, and it is worth saying plainly that it makes the engineering problem harder, not easier. Choosing the less industrialised region is a choice to pay more, in time and in infrastructure, for a distributional outcome. Stated openly, it is defensible. Folded into the word speed, it simply becomes a cost that shows up long after the announcement.
The timing is not an accident either. Samsung and SK hynix are the world’s two largest producers of memory for AI applications, and they are riding a surge in global demand for high-performance chips, HBM memory in particular. Committing sums of this size while demand for your core product is climbing is not courage, it is timing. It is also worth remembering that announced investment is the easiest kind to stretch out, phase differently or quietly reschedule if conditions change, which is exactly what makes the 2029 and 2035 dates the only interesting numbers in the whole package.
So the plan will not be judged by the size of the figure or the speed of the signatures. It will be judged where the two speeds meet. If the authorisations arrive on schedule and the water supply, the infrastructure and the trained workforce do not, the fast half of the plan spends years waiting on the slow half. Money has been announced for the slow half too, infrastructure included. Turning those authorisations into operating capacity is the work that still has to be done, and that work keeps its own calendar.
Raffaele Di Marzio
All my “insane” books on cybersecurity and governance are here 👇 https://cyberium.limited/bookshelf.html