South Korea’s new megaproject plan is easy to misread as another semiconductor expansion story. The headline numbers are dominated by memory fabs, HBM packaging, and AI data centers. But the larger plan is more ambitious: it is an attempt to redraw Korea’s industrial geography around artificial intelligence.
The framework began with the June 29 “Three Megaprojects for a Great Leap Forward” briefing, where the government outlined three connected pillars: semiconductors, physical AI, and AI data centers. The plan links memory-chip capacity, robotics, manufacturing AI, and large-scale compute infrastructure into one national industrial strategy. In the government’s own framing, these are not separate projects, but a “triangle” for Korea’s next phase of industrial competitiveness.
That makes the regional map as important as the technology itself. Honam and the southwest are being positioned as Korea’s second memory semiconductor production base. Chungcheong is being developed as a hub for HBM, advanced packaging, displays, batteries, and bio manufacturing. Yeongnam and Daegyeong are being assigned a different role: physical AI, robotics, materials, components, equipment, power semiconductors, and manufacturing conversion. AI data centers are intended to connect these regional clusters into a national compute network.
The result is not simply “Samsung and SK Hynix will build more fabs.” It is a proposed reorganization of Korea’s industrial base for the AI era.
Honam and the southwest: Korea’s second memory base
The most visible part of the plan is the southwest semiconductor push. The government announced that the southwest would receive an 800 trillion won semiconductor investment package, including four memory fabs and the supplier and workforce ecosystem needed to support them. The stated goal is to expand Korea’s memory production capacity and reduce pressure on the already crowded Seoul-Gyeonggi semiconductor corridor.
A follow-up southwest briefing on June 30 added more detail. SK Hynix, Samsung Electronics, and Amkor signed a memorandum of understanding with the government for a combined 896 trillion won in planned investment. SK’s portion was described as about 470 trillion won for two memory main fabs and a 1GW AI data center. Samsung’s portion was described as 425 trillion won for two memory fabs and a national AI computing center in Honam. Amkor said it would invest 1 trillion won to expand an advanced packaging facility in Gwangju.
This gives Honam a central role in the plan. The region is not being treated only as a beneficiary of balanced regional development. It is being positioned as the second lung of Korea’s semiconductor industry, alongside the existing capital-area production base.
The logic is practical. Advanced memory demand is being driven by AI servers, high-bandwidth memory, data-center infrastructure, and future AI devices. But chip production needs land, power, water, logistics, specialized suppliers, and skilled labor. The southwest plan is therefore as much an infrastructure plan as it is a semiconductor plan.
That is also why many details remain unfinished. The investment figures are planned commitments and MOU-based announcements, not completed construction. Exact site decisions, construction schedules, power-grid upgrades, water-supply plans, supplier relocation, and workforce training will determine whether the southwest can become a real second semiconductor base rather than a political slogan.
Chungcheong: the HBM and advanced-industry bridge
Chungcheong’s role is different. It is not simply another fab region. It is being developed as the bridge between memory production, advanced packaging, displays, batteries, and bio manufacturing.
At the June 29 briefing, the government said 81 trillion won would be invested to make Chungcheong a semiconductor packaging hub. The plan includes new HBM-related facilities in Onyang and Cheonan and HBM packaging investment in Cheongju. That matters because the AI semiconductor race is no longer only about how many wafers can be produced. It is also about how memory is stacked, packaged, connected, cooled, and integrated into AI accelerators and servers.
The July 2 Chungcheong briefing widened that picture. The government said it had attracted 392 trillion won in corporate investment for the region. Samsung Display announced about 140 trillion won for OLED, next-generation display lines, HBM fabs, and packaging. SK Hynix announced about 100 trillion won for NAND flash production facilities and advanced packaging fabs. Celltrion announced about 2 trillion won for biopharmaceutical production facilities. The companies and central and regional governments also signed an MOU to support the investment.
This makes Chungcheong the plan’s high-value processing zone. Honam is being framed around new memory production capacity. Chungcheong is being framed around the technologies that help turn memory and adjacent advanced industries into higher-value systems.
The region also has a broader strategic identity. The government has emphasized that semiconductors, displays, batteries, and bio are four core advanced industries for Korea’s AI-era economy. Chungcheong is unusual because all four are already present or being expanded in the same regional belt.
Yeongnam and Daegyeong: physical AI and manufacturing conversion
If Honam is the memory base and Chungcheong is the advanced-packaging and processing hub, Yeongnam and Daegyeong are being positioned as the physical AI and manufacturing-conversion base.
The June 29 national plan assigned the southeast and Daegyeong regions roles in materials, components, and equipment for the semiconductor supply chain, as well as future semiconductor fields such as power semiconductors. The same plan also tied Daegyeong’s existing auto-parts and home-appliance supplier base to the development of robotics and physical AI.
This is an important distinction. Yeongnam is not being framed primarily as the next big memory-fab region. Its role is closer to the “body” of the AI economy: robots, sensors, actuators, robot hands, power devices, factory automation, industrial equipment, defense-related chips, and AI-enabled manufacturing systems.
That role fits the region’s industrial base. The southeast already has deep manufacturing capacity in autos, shipbuilding, machinery, energy, aerospace, and components. The government’s plan is to move those legacy strengths into AI-era production: smarter factories, humanoid and industrial robots, robot parts, energy systems, and next-generation manufacturing equipment.
This part of the plan is still less concrete than the southwest and Chungcheong packages. The broad role is clear in the national framework, but the most important implementation details—specific sites, company-by-company investment schedules, supply-chain conversion programs, and local workforce plans—are still the pieces to watch.
AI data centers: the national compute layer
The third pillar is AI data centers. This is where the semiconductor plan and the physical-AI plan connect.
The government said SK, GS, and Naver would work with the state to build a first-phase AI data-center network of 8.4GW by 2029: 5GW from SK, 2.4GW from GS, and 1GW from Naver. The first-phase investment was described as about 550 trillion won, including investment attraction. SK also presented a second-stage expansion plan that would bring its capacity from 5GW to 15GW by 2035.
These data centers are not just server warehouses. In the government’s plan, they are testbeds for Korean AI semiconductors, power systems, cooling systems, cloud tools, and data-center solutions. If successful, Korea would not only build AI data centers domestically; it would try to export the technologies and operating models behind them.
This also explains why power and water appear throughout the plan. AI data centers and semiconductor fabs are both infrastructure-intensive. The government has discussed using renewable energy, nuclear power, expanded transmission networks, reused water, multipurpose dams, and faster permitting to support the projects. It has also floated dedicated electricity-pricing systems for AI data centers.
In other words, the megaproject is also an energy and infrastructure policy.
Manufacturing AI: the layer that ties the plan together
The most important connective idea is manufacturing AI. The government’s physical-AI strategy includes accelerating manufacturing AI, building data factories across major industries, developing specialized AI robots, and deploying more than 1,000 AI robots per year into industrial sites. It also calls for domestic development of robot components and training 10,000 AI robotics specialists over five years.
This is where the plan becomes more than a list of investments. Memory chips support AI systems. AI data centers provide compute. Manufacturing AI improves factories. Robots and physical AI bring intelligence into industrial settings. Regional clusters supply the components, equipment, workers, and production sites.
The strategy is circular: AI infrastructure supports manufacturing, manufacturing generates data and demand, and that demand feeds back into semiconductors, data centers, robots, and industrial software.
What has been announced, what is in progress, and what remains uncertain
The announced pieces are significant. The national framework has been presented. The southwest and Chungcheong briefings have produced company-level investment announcements and MOUs. The regional roles are now visible: Honam for memory fabs and AI compute, Chungcheong for HBM and advanced industries, Yeongnam and Daegyeong for physical AI and manufacturing conversion, and a national layer of AI data centers connecting them.
But the plan should still be read carefully. Many of the numbers are planned investments, investment roadmaps, or MOU-based commitments. They are not the same as completed spending. The difference matters, especially because semiconductor fabs and AI data centers require long permitting cycles, enormous power supply, reliable water access, supplier ecosystems, housing, transportation, and thousands of specialized workers.
The highest-risk areas are clear. First is infrastructure. Without power, water, roads, rail, and industrial land, even the largest corporate commitment cannot become a working fab or data center. Second is workforce. Korea will need engineers, technicians, construction workers, data-center operators, robotics specialists, packaging experts, and AI manufacturing talent outside the existing Seoul-Gyeonggi talent pool. Third is supplier density. Semiconductor clusters work because suppliers, equipment companies, logistics providers, universities, and research institutes operate close together. Building that ecosystem in new regions will take time.
The fourth risk is execution speed. The government says it wants to shorten industrial-site development and permitting timelines, but large Korean industrial projects have often moved slowly because of land, environmental review, local opposition, grid constraints, and inter-ministerial coordination.
That is why the complete picture is both ambitious and unresolved. South Korea has announced a national AI-industrial map. It has not yet proven that the map can become a functioning industrial system.
The bigger meaning of the plan
The megaproject plan shows how Korea is thinking about the next stage of global competition. The country is not only trying to defend its memory-chip lead. It is trying to connect memory, HBM, data centers, robotics, manufacturing AI, batteries, displays, bio, aerospace, and energy into one national growth model.
Honam is being asked to become the new memory production base. Chungcheong is being asked to become the advanced-packaging and high-value industry bridge. Yeongnam and Daegyeong are being asked to convert their manufacturing depth into physical AI and robotics strength. AI data centers are being asked to provide the compute layer that makes the system work.
That is why the plan is bigger than chips. If it succeeds, it could reshape Korea’s industrial geography for the AI era. If it stalls, the biggest obstacles will not be ambition or corporate branding. They will be land, power, water, labor, suppliers, and execution.
