South Korea’s $576bn chipmaking expansion plan places the country’s two dominant memory manufacturers at the centre of a wager that AI infrastructure spending will still be accelerating when the first new fab comes online, which won’t be before 2028 at the earliest.
President Lee Jae Myung announced today that Samsung Electronics and SK Hynix will invest 800 trillion won ($518bn) alongside local governments to build new chipmaking facilities in the country’s south-western region, with a further 81 trillion won directed toward a chip-packaging cluster in Chungcheong, near Seoul. The investment is part of Lee’s broader “three megaprojects” initiative, targeting semiconductors, physical AI, and data centres. South Korea’s science minister separately announced that the country is targeting 550 trillion won in AI data centre investment by 2029 and more than 1,000 trillion won by 2035.
After markets closed on Monday, Samsung went further. The company announced its largest-ever domestic investment plan, totalling up to 2,655 trillion won, exceeding market expectations of around 2,000 trillion won. More than 76% of those funds, roughly 2,030 trillion won, are earmarked for Samsung’s existing Pyeongtaek campus and the Yongin National Industrial Park, focused on expanding advanced foundry processes and building capacity for HBM4, HBM5 and next-generation memory. The remaining funds are distributed across the south-western region and other sectors. That after-hours announcement reframes the Blue House event: the joint government-industry figure was the headline, but Samsung’s own commitment is considerably larger and more specifically targeted at reclaiming its competitive position in HBM.
The scale of both announcements is designed to be read as a geopolitical signal as much as an industrial one. Both companies are already building facilities in the United States under pressure from President Donald Trump’s semiconductor reshoring agenda. Monday’s announcements are Seoul’s domestic counterpart: a statement that the country’s manufacturing base isn’t migrating, it’s expanding. Together, Samsung and SK Hynix hold roughly 67% of the global DRAM market by revenue, with Samsung at 38% and SK Hynix at 29%, according to Counterpoint data from the first quarter of 2026. That grip on supply has made them central to the global buildout of AI infrastructure, since HBM chips, which are produced by stacking multiple DRAM layers, are currently the most constrained component in the AI server supply chain.
The problem with building for a constrained market is that the constraint is exactly what this investment will eventually resolve. Chip factory construction operates on timelines that don’t align neatly with demand cycles. Samsung Electronics Chairman Jay Y. Lee confirmed at Monday’s event that the company has selected Gwangju as a site for its new cluster, with experts proposing several potential sites including the grounds of a military air base slated for relocation. SK Hynix’s Chairman Chey Tae-won offered a more telling detail: his firm still needed additional time to finalise a site and secure infrastructure in the south-western region. “It took us nine years for us to create a cluster in Yongin,” Chey said. “Also, a chip factory requires massive land, power, water and talent.”
That’s not a minor caveat. Industry experts warn that building cutting-edge fabs requires vast electricity and water, advanced logistics, deep supplier networks, and highly skilled labour, none of which scales quickly in a greenfield region. Kim Tae-yun, a professor of administration at Hanyang University, said securing skilled workers will be extremely difficult in the southwest, and that will determine whether the project succeeds or fails. “Unless a truly cutting-edge fab is built, the local economic impact will be limited, it risks becoming little more than a construction project and a real estate boost,” he said. Government officials dismissed questions about whether the southwest has enough power and water to support major semiconductor fabs, saying the region’s strength in renewable energy would give the chipmakers an edge as they face growing global pressure to use cleaner sources of electricity.
To address the execution timeline directly, South Korea’s minister of trade, industry and resources, Jung-Kwan Kim, said the government will rapidly expand production capacity by drastically shortening the timeline from licensing to construction. That’s an acknowledgement that the multi-year permitting and build-out process is a real bottleneck, though shortening regulatory timelines and solving for skilled labour in a new region are different problems.
Lee Jong-ho, a professor at Seoul National University’s Department of Electrical and Computer Engineering, put the underlying demand risk plainly: “It would be ideal if demand remained strong for the next 20 or 30 years, but no one can know that with certainty. If demand were to decline, the consequences would be severe.”
SK Hynix only recently overtook Samsung as South Korea’s most valuable publicly traded company, reaching a market capitalisation of approximately 1.35 trillion won in mid-June 2026, driven by explosive demand for HBM chips that power AI training and inference. Samsung’s 2,655 trillion won after-hours commitment reads partly as a statement of intent to close that gap, with HBM4 and HBM5 capacity central to the plan. The competitive dynamic between the two companies, as much as the AI demand story, explains the scale of what was announced on Monday.
Markets weren’t persuaded on the day. Samsung shares closed down 4.86% while SK Hynix fell 1.68%, with analysts flagging potential oversupply risk. Morningstar analyst Jing Jie Yu noted that new capacity takes two to three years at minimum to come online and historically arrives just as peak demand begins to taper. The history of chip manufacturing is littered with examples of overcapacity following periods of aggressive expansion, and the involvement of chaebols introduces governance considerations that foreign investors tend to scrutinise, given their concentrated ownership structures and complex cross-holdings.
There’s also a political dimension the reporting makes explicit. Opposition politicians have sharply criticised the southwest chip hub, questioning whether the location was chosen on industrial merit or political convenience, given that 85% of voters in the region backed Lee in last year’s presidential election. The president spent the weekend defending the plan on X before Monday’s announcement. The main opposition People Power Party said semiconductor factory locations should be decided by companies, not by the president.
None of this makes the investment obviously wrong. The demand is real, the shortage is real, and the existing sites around Yongin and Pyeongtaek have, by Lee’s own assessment, reached their geographic limits. The competitive pressure from China adds urgency too. Reuters reported separately this week that China’s CXMT has secured a $3bn memory supply deal with Tencent, a signal that Chinese memory producers are gaining ground in domestic enterprise markets even while they remain locked out of leading-edge HBM production. South Korea has every incentive to expand capacity before that gap narrows.
For South Africa, the relevance is indirect but not trivial. Local data centre investment has accelerated sharply, with hyperscaler commitments from Microsoft, AWS, and Google reshaping infrastructure across Gauteng and the Western Cape. All of that capacity depends on memory chips whose price and availability flow directly from decisions made in Seoul. When Apple raised MacBook and iPad prices last week, citing unsustainable memory and storage costs, it illustrated precisely how constrained supply translates across the value chain and reaches every South African business sourcing hardware to run AI workloads. As we’ve covered in our analysis of Google’s AI inference infrastructure push, the physical layer of AI is where access and cost are actually determined, and memory remains the tightest constraint in that layer.
The $576bn government figure and Samsung’s own 2,655 trillion won after-hours commitment will be quoted widely as evidence of South Korea’s semiconductor ambition. The more useful measure is whether the demand that justifies both holds when the fabs are ready. SK Hynix’s chairman effectively acknowledged on Monday that nobody, including the companies making these investments, can answer that with confidence. That’s not a reason to walk away. It’s a reason to understand what the bet actually is.