SK Hynix's Chairman Says Build Faster: Korea's $300 Billion Bet on the Future of Memory

On October 11, 2026, SK Group chairman Chey Tae-won stood before business leaders in Shanghai and delivered an unusually blunt message: South Korea must build semiconductor factories faster, or watch rivals take the market share its memory champions spent decades earning. The warning came two days after Chey walked a sprawling field outside Gwangju in southwestern South Korea, the future home of a planned semiconductor cluster on the site of a military airport, where he signaled that SK Hynix could construct new fabs in parallel with its already accelerated Yongin project instead of treating the two as sequential bets.

The message is striking because it comes from a company that is already moving at a pace the industry barely recognizes. SK Hynix has brought forward the planned completion of its Yongin semiconductor cluster by 12 years, from 2045 to 2033. Chey's point is that even this may not be enough.

Why Korea is suddenly in a hurry

Chey named the competitors keeping him awake. On one side is Micron Technology, the U.S. memory producer adding capacity at home. On the other are Chinese chipmakers ChangXin Memory Technologies (CXMT), which produces DRAM, and Yangtze Memory Technologies Corp. (YMTC), which produces NAND flash. Chey argues both are increasing their production capabilities and technical strength at a rapid pace, and that Korean producers cannot count on technology leadership alone to protect their position.

His competitive logic is about timing, not just technology. In his telling, strong demand for memory does not automatically turn into sales for Korean companies if their factories are already running near capacity: competitors that bring new capacity online sooner can capture the orders Korean producers cannot yet supply. An SK Hynix official put the demand picture starkly, saying AI memory demand is exploding, market growth is accelerating, and competition is becoming increasingly fierce. Chey has previously said memory production capacity must more than double within five years to answer the AI era.

The expansion debate is complicated by Washington. John Moolenaar, chairman of the U.S. House Select Committee on Strategic Competition between the United States and China, recently wrote to Secretary of State Marco Rubio warning about Chinese factories operated by Samsung Electronics and SK Hynix, specifically Samsung's Xi'an facility and SK Hynix's Wuxi and Dalian facilities, which he argued could create vulnerabilities in U.S. supply chains. Chey rejected the premise that restricting those operations would solve the underlying problem, arguing that blocking Korean companies' plants in China will not stop China's own chipmakers from advancing. The letter is a congressional intervention, not a regulation: no new policy decision or deadline has resulted, but the uncertainty hangs over Korean investment planning.

Chey's message to Korea's semiconductor industry is simple: factory speed, not technology alone, will decide who supplies the AI boom.

The Honam blueprint

The centerpiece of the urgency campaign is the Honam semiconductor cluster near Gwangju. According to Reuters reporting from October 9, Samsung Electronics and SK Hynix each plan to invest about 400 trillion won, roughly $298 billion at recent exchange rates, to build two chip fabs apiece in the cluster. The site spans about 8.26 million square meters across Gwangju's military airport, which is due to be relocated, and a neighboring ammunition depot. Together, the two companies' commitments form part of the government's broader industrial plan, sometimes described as three mega projects, totaling about 800 trillion won for four new memory fabs.

SK Hynix unveiled its Honam blueprint in June, and company executives have since visited the Gwangju site more than ten times to coordinate with local authorities. Chey described the planned scale as about two-thirds that of the Yongin cluster, with room to grow if more land and demand become available. He declined to disclose a specific number of fabs, but local authorities are targeting construction to begin in early 2028, initial water and power infrastructure by the end of 2028, and mass production in 2030. The sequencing is deliberate: Chey said SK is willing to begin construction work such as piling even if utility delivery lags slightly, provided public infrastructure will be ready when manufacturing equipment actually needs it. He added that he has no intention of cutting construction costs to make the numbers work.

The infrastructure demands are enormous. South Korea's climate, energy and environment ministry has pledged 6.3 gigawatts of power and 650,000 tons of water for the project, and the government is seeking to exempt the water and electricity infrastructure from preliminary feasibility studies to speed things along. Part of the funding would come through a future response fund financed by excess tax revenue from the semiconductor industry. One structural advantage: the Honam region is rich in solar and offshore wind resources, which helps with RE100 renewable energy requirements that global tech customers increasingly demand.

Money is already moving

The urgency is showing up in real spending, not just speeches. On October 8, SK Hynix revised its planned fourth-quarter 2026 transactions with affiliate SK ecoplant for fab construction and supporting production equipment to 4.25 trillion won, a jump of at least 20 percent over the figure disclosed in February. The concrete is the message: SK is accelerating actual construction spending, not merely announcing future intentions.

That spending is politically contested. Opposition lawmakers have challenged the allocation of the future response fund, arguing it is too heavily concentrated on the Honam project and questioning whether it serves regional political interests. The planning and budget minister responded that the spending is strategic investment meant to sustain the semiconductor industry, not a regional handout. A parliamentary audit scheduled for October 22 in the Jeonnam and Gwangju area will be the next occasion for scrutiny of site selection, infrastructure financing, and whether the public commitments hold. For SK, the political debate matters directly: the speed of factory construction depends partly on public infrastructure whose financing and execution remain contested.

The demand engine behind the race

Underneath the politics sits a demand forecast that justifies the scale. Market research firm Omdia projects the global memory market will roughly quadruple from about $200 billion this year to around $800 billion by 2030, driven by the race to build AI data centers. The anxiety is not only about missing revenue: analysts warn that if a large-scale shortage is prolonged, customers could accelerate research into alternative technologies, and the memory industry could permanently lose part of its addressable market. Undersupply carries its own penalty.

SK is hedging geographically while keeping the domestic build as the priority. Chey confirmed that overseas factory expansion, including in Japan, remains under study but is not a priority decision, and he used the Shanghai conference to announce plans for an AI research and development center in the city covering AI data centers, physical AI, and biopharmaceuticals. He noted the group has invested about $60 billion in China over 35 years across semiconductors, batteries, and petrochemicals. The near-term test, though, is closer to home: converting the Honam commitments, the accelerated Yongin schedule, and a politically scrutinized infrastructure plan into factories that can hit mass production in 2030. The timeline leaves almost no margin for delay, which is exactly Chey's point.