South Korea is spending $880 billion on chips, AI, and robots over the next decade

South Korea announced on June 28 that it will deploy 1,350 trillion won, roughly $880 billion at current exchange rates, into semiconductors, AI data centres, and physical AI and robotics over the next ten years. The figure is not a projection or an aspiration. Samsung and SK Hynix, the two companies that dominate global memory chip production, are anchoring the plan with four new fabrication plants in the country's southwest.

For scale: $880 billion is approximately 5% of South Korea's entire 2024 economic output. It is more than double the US CHIPS Act. It is more than any single country has committed to semiconductor investment in history. President Lee Jae-myung called the three investment pillars, chips, data centres, and robotics, a "triple axis" of future technology.

What the money is actually buying

The plan breaks into three categories. The 800 trillion won manufacturing commitment is for semiconductor fabrication, concentrated in four new fabs in South Korea's southwestern provinces. That geographic choice is deliberate: the new plants are being placed in areas that have historically been left behind by Seoul's economic concentration, using the AI semiconductor boom as a regional development tool.

The 550 trillion won data centre commitment comes from companies including Naver, South Korea's dominant internet company, with a target of 8.4 gigawatts of AI data-centre capacity by 2029. For context, the entire US AI data centre build-out over the last two years sits at roughly 20 gigawatts. South Korea is targeting 8.4 gigawatts from a standing start in three years.

The robotics component covers physical AI systems, humanoid and industrial, reflecting South Korea's ambitions in hardware manufacturing beyond chips.

Why memory chips are the key bet

South Korea's semiconductor speciality is memory: DRAM and NAND flash. Samsung is the world's largest memory chip maker. SK Hynix makes the high-bandwidth memory (HBM) that goes into Nvidia's H100 GPUs, making it a critical node in the AI supply chain whether Nvidia retains dominance or not.

Memory is increasingly the bottleneck in AI inference. The chips doing the computation are powerful enough; getting data to and from them fast enough is the constraint. HBM solves part of that problem, and SK Hynix is in a multi-year partnership with Nvidia to co-develop next-generation memory for AI supercomputers. That partnership is now being backed by an $880 billion national commitment.

Where this sits in the global chip race

Taiwan's TSMC remains the world's leading manufacturer of advanced logic chips, the kind that go into Apple processors, Nvidia GPUs, and AMD CPUs. South Korea is not trying to compete with TSMC on logic; it is betting that memory becomes the defining constraint in AI infrastructure, and that whoever controls the best memory chips at scale controls a critical node in the whole AI supply chain.

The US CHIPS Act allocated roughly $52 billion to domestic semiconductor manufacturing. South Korea's plan is $880 billion. The EU Chips Act committed 43 billion euros. Japan's semiconductor subsidies have totalled around $13 billion. The investment gap between South Korea and everyone else is stark.

The question is whether the fabs come online on schedule. South Korea has a strong track record in semiconductor manufacturing execution. If these plants deliver, they shift the global chip production map meaningfully by the early 2030s.