Related: Big Tech’s 2.7 trillion dollar AI bill comes due.
South Korea has placed one of the largest national bets yet on artificial intelligence, and it is a bet on hardware rather than software. Standing alongside the leaders of Samsung Electronics and SK Hynix, President Lee Jae-myung unveiled a plan to steer roughly 880 billion dollars of public and private investment over the next decade into semiconductors, AI data centers, and robotics. The message behind the enormous number was direct, that the country which already makes the memory inside the world’s AI chips intends to turn that position into lasting dominance.
Lee framed the effort around what he called a triple axis of chips, physical AI, and data centers, and the centerpiece is a wave of new factories. Samsung and SK Hynix will together build four new chip fabrication plants, two each, in the country’s southwest, a combined commitment of around 800 trillion won, or roughly 518 billion dollars, aimed squarely at expanding memory production. Samsung’s share includes hundreds of billions of dollars for new fabs in the Gwangju area, plus additional lines in Cheonan and Onyang dedicated to advanced high bandwidth memory, the specialized chips that AI accelerators depend on.
The reason the world should pay attention sits in that last detail. Samsung and SK Hynix produce most of the high bandwidth memory, known as HBM, that is stacked alongside the processors inside modern AI systems, and demand for it has raced far ahead of supply, with the waitlist stretching into 2027. Every company building large AI models needs those memory chips, and there are only a handful of firms on earth that can make them at the required quality. By pouring money into more capacity, South Korea is not just chasing its own growth, it is positioning itself as an indispensable supplier in a market where scarcity has become the defining constraint.
The scale of the commitment is hard to overstate in national terms. The total planned spending amounts to roughly 5 percent of the country’s entire economic output for 2024, an extraordinary share to direct at a single industrial push. Reported figures for the plan have varied depending on what is counted, from the 518 billion dollars in chip plants to the broader 880 billion dollar total that folds in data centers, robotics, and other technology firms, but the direction is consistent across every version. This is a whole of country effort to make semiconductors and AI infrastructure the engine of the next decade of Korean growth.
Beyond the fabs, the plan reaches into the infrastructure that AI runs on. It calls for a massive expansion of AI data centers, with reporting pointing to a target of more than 18 gigawatts of capacity by the mid 2030s, alongside investment in next generation memory research and the robotics that Lee grouped under the banner of physical AI. The logic is that leading in chips alone is not enough if the country cannot also build the data centers to deploy them and the automated systems to apply them, so the plan tries to strengthen the entire chain at once rather than a single link.
The gamble is not without risk. Committing a sum equal to a large slice of the economy to one sector ties the country’s fortunes tightly to the continued growth of AI demand, and a slowdown, or a shift in the technology that reduced the need for ever more memory, would land hard. Building fabs is also slow and staggeringly expensive, and the payoff will unfold over years rather than quarters. But Lee was blunt about the stakes, telling executives that South Korea must secure the core elements of AI faster than any rival or risk being left behind. In a global contest increasingly defined by who controls the physical supply of computing, South Korea has decided that the safest move is to spend its way to the front and stay there.
Your First 10 AI Skills
10 practical AI skills, copy-paste prompts and a 7-day plan to start using AI with confidence.
Download the guide →