South Korea is betting over $1 trillion — about 1,350 trillion won — on a three-decade transformation built on memory chips, AI data centers, and commercial humanoid robots, with a government-backed program announced June 29, 2026 that puts it at the center of the global AI infrastructure race. The announcement, made by President Lee Jae Myung, maps directly onto what tech giants call “RAMageddon” — a worldwide shortage of high-bandwidth memory that every major AI model now depends on.
Cover image: South Korea memory chips, data centers and humanoid robots investment. Photo credit:Jeon Han / CC-BY 2.0.
The Three-Pillar Plan Behind Korea’s $1 Trillion Wager
This is not a single industry initiative. Seoul structured the investment as three national mega-projects: semiconductors, AI data centers, and physical AI — a term it uses for humanoid robots and autonomous systems. Samsung Electronics and SK Hynix are anchoring the semiconductor half, while Naver and the major telecom carriers are backing the data-center push. The humanoid robot track, targeted for commercial deployment by 2028, is the wild card, and it is the piece that has drawn the most international attention.
The plan mirrors the logic of the U.S. CHIPS Act, but on a much larger dollar scale. It also reflects a clear-eyed Korean assessment: if AI models are eating memory faster than anyone can manufacture it, the country that controls DRAM and HBM production controls the pace of AI progress itself.
Memory Chips: The Immediate Battle Against ‘RAMageddon’
The chip portion of the program dwarfs everything else. Samsung, SK Hynix, and their supply-chain partners are committing 800 trillion won — roughly $518 billion — to build four new memory fabrication facilities in South Korea’s southwest. The sites include Gwangju and the Honam region, plus Onyang and Cheonan in Chungcheong. A separate 81 trillion won HBM packaging hub will rise in Cheongju.
SK Hynix, which in early 2026 became South Korea’s most valuable company for the first time in 26 years, is targeting one million monthly DRAM units by 2030 and tripling wafer capacity by 2034. Samsung is planning a roughly 50% HBM capacity increase in 2026 alone. Those numbers are not incremental upgrades — they represent a doubling of Korea’s semiconductor footprint inside a decade.
Samsung and SK Hynix Lead the Charge
The rivalry between Samsung and SK Hynix is now a strategic asset for Seoul. Both companies are expanding HBM and DDR5 production simultaneously, and both are racing to meet commitments to hyperscalers that have pre-bought memory years in advance. SK Hynix’s rise over Samsung marks a shifting power dynamic inside Korea’s own semiconductor ecosystem, and both firms are now sitting on backlogs that make the $518 billion spend easier to finance.
The shortage is also reshaping other industries. Smartphone shipments are on track for a 15% decline in 2026 because memory prices are climbing so fast that end-product costs are becoming unsustainable. Korea is trying to short-circuit that cycle by flooding the zone with new supply.
AI Data Centers: Building 8.4 GW of Infrastructure
The data-center side of the plan calls for 8.4 gigawatts of AI-optimized capacity across three new sites. Naver and Korea’s major telecom carriers are the stated partners. The government says it will accelerate GPU acquisition, with reports suggesting 18,000 GPUs delivered in the first half of 2026 alone.
Eight-point-four gigawatts is not an abstract metric. For comparison, a typical large hyperscale data center campus consumes 150–300 megawatts. Korea is effectively planning enough capacity to run dozens of the world’s biggest AI compute facilities. That scale commits the country to being a permanent landlord for AI training workloads that will not finish running before 2040.
Humanoid Robots: Korea’s 2028 Commercial Target
The humanoid-robot timeline is the plan’s most ambitious and least certain element. Seoul wants commercial humanoid robot deployment by 2028. Hyundai Motor Group and Samsung ecosystem companies are expected to play lead roles. Jensen Huang visited Seoul in June 2026 and told local audiences that robotics was Korea’s next major AI export sector.
The “physical AI” framing matters here. Unlike software models, humanoid robots require mechanical engineering, sensor fusion, and battery breakthroughs on top of AI inference. The race from pilot projects to mass production is already intense globally, and Korea’s entry gives the category its first true manufacturing superpower. Chip advances such as IBM’s sub-1nm Nanostack show how computing density and robotics hardware are advancing in parallel.
Why This Changes the Global Tech Map
Korea’s plan echoes a broader East Asian semiconductor arms race, but its combination of memory dominance, committed data-center buildout, and a forced humanoid-robot deadline makes it structurally different from Japan’s earlier chip bets or Taiwan’s foundry-centric approach. Memory chips — specifically high-bandwidth memory — remain irreplaceable in AI training, which makes SK Hynix’s HBM leadership a strategic chokepoint the U.S., Europe, and China are all desperate to avoid.
If execution holds, the program could double Korea’s semiconductor manufacturing footprint inside 20 years and create a domestic supply chain for humanoid robots that rivals anything Detroit or Tokyo can offer. If any piece slips, the economics of the entire bet become harder to sustain. Either way, the announcement has already reshaped the global AI infrastructure conversation.
Watch: South Korea’s $520 billion-plus memory and AI chip investment
Briefing and analysis covering the government-backed rollout and its likely impact on global memory supply.
FAQ
Is South Korea’s $1 trillion investment already guaranteed funding?
No. The June 29, 2026 announcement is a national commitment backed by government encouragement and stated corporate intent from Samsung and SK Hynix. Actual spending will depend on market conditions, construction timelines, and capital allocation decisions over the coming decade.
Why is memory chip production so critical for AI?
AI training workloads rely on high-bandwidth memory (HBM) and advanced DRAM to store model parameters and intermediate calculations. As models grow larger, memory becomes the bottleneck. SK Hynix and Samsung together supply the majority of the world’s HBM, making Korean output central to every major AI lab’s roadmap.
When will Korea’s humanoid robots actually reach the market?
The program targets commercial humanoid robot deployment by 2028. This is an aggressive goal, and the technology is advancing quickly, but mass-market robotics at scale has historically missed timelines. 2028 is the stated target, not a guarantee.
Facebook Twitter LinkedInWhatsAppThreads Telegram Email
CopyCopied
Tags:
AIAI InfrastructureHBMhumanoid robotsMemory chipssamsungSK HynixSouth Korea
Originally published on TekMag
Top comments (0)