South Korea has decided the next phase of the AI race will be won in factories, not chatbots. On 30 June, President Lee Jae-myung unveiled a plan to pour roughly $880 billion into semiconductors, AI data centres and robotics over the coming decade, a commitment his government is calling the "Three Mega Projects" and framing in the language of national survival.

The largest single piece goes to chipmaking. Samsung and SK Hynix, together with their suppliers, will invest about 800 trillion won, close to $518 billion, to build new fabrication plants in the country's southwestern Honam region, according to PBS. The industry ministry says the goal is to double the country's memory-chip production capacity within five years and hold its lead against rising competition from China and Taiwan.

A second tranche, around 550 trillion won or $356 billion, will fund AI data centres backed by the SK Group, GS Group and Naver. Looking further out, Seoul wants an additional 10-gigawatt data centre online by 2035, a build-out that pushes the full tally past a trillion dollars once the later phases are counted.

A bet on hardware

What makes the announcement striking is where the money is going. While much of the industry spends on models and the talent to train them, South Korea is betting that whoever controls the memory, the fabs and the power-hungry data centres will hold the real leverage. Bloomberg framed it as Lee tying his political legacy to the fortunes of the domestic chip boom, which is a sizeable wager for any leader to make.

The logic is not hard to follow. High-bandwidth memory has become one of the genuine chokepoints in AI hardware, and Samsung and SK Hynix already sit near the centre of that market. Doubling capacity is an attempt to make the shortage a Korean advantage rather than a Korean constraint.

The catch

Numbers this large invite scepticism, and there are fair reasons for it. Much of the $880 billion is projected private investment rather than committed state spending, and a decade is long enough for demand forecasts to shift more than once. The plan also concentrates enormous construction in a single region, hours from Seoul, which raises real questions about power supply, water and the skilled workforce needed to run the sites.

Still, the direction is unmistakable. Korea is joining a scramble for compute infrastructure that has already produced OpenAI's move into custom silicon with Broadcom and Qualcomm's push to challenge Nvidia's stack. The difference is scale. This is a whole country, with its two largest companies, deciding the physical layer of AI is worth the better part of a trillion dollars.

Sources

  1. i. www.bloomberg.com
  2. ii. www.pbs.org
  3. iii. www.techspot.com

Commentarii · 0

Add · a · Comment