China's Semiconductor Rise and Korea's Response: A Focus on Memory Chips

Byung-Chul LEE

bcclee65@naver.com​

Visiting Research Fellow

Sejong Institute

1. Introduction

Semiconductors are the core pillar of the Korean economy and its leading strategic industry with a global competitive edge, as well as a national strategic asset of great importance for economic security. In the first half of 2026, semiconductors accounted for 38.7% of Korea's total exports, and in June 2026 the combined market capitalization of Samsung Electronics and SK hynix exceeded half of the entire KOSPI. Korea's exports and capital markets are thus heavily dependent on the semiconductor industry. With Korea's advantage in shipbuilding, steel and displays already narrowing, the core problem is that no industry could replace semiconductors in the short term should they falter.

It is no longer sufficient to assess the threat from Chinese semiconductors by asking how many years China lags behind technologically. China still trails considerably in advanced segments such as HBM (high-bandwidth memory) and advanced server DRAM. In NAND flash and commodity DRAM, however, it is rapidly expanding its market share and entering a stage where it can influence the global market. For now, strong AI demand and memory supply shortages are absorbing the effects of China's capacity expansion, keeping the pressure largely out of sight; once demand slows, the impact of this expanded supply could come fully to the surface.

The United States is using supply-chain chokepoints such as advanced equipment, AI chips and EDA (electronic design automation) software to restrict China's access to advanced technology. In response, China is accelerating localization and the building of an independent technology ecosystem under the banners of "self-reliance and controllability" (zizhu kekong, 自主可控) and Xinchuang (信创, IT application innovation). A technology-security dilemma is deepening: US controls spur China's technological self-reliance, and China's self-reliance in turn invites further controls.

Against this backdrop, this paper examines how far Chinese semiconductors have actually caught up in technology, production capacity and markets, under what conditions the threat to Korean semiconductors will become acute, and what Korea should do.

2. The Global Semiconductor Industry and Competitive Landscape

The global semiconductor market is growing rapidly on the back of expanding AI investment, with memory recently leading the expansion. Estimates based on WSTS data suggest that the global semiconductor market will grow about 2.6-fold, from US$630.5 billion in 2024 to about US$1.655 trillion in 2026. Over the same period, the memory market is expected to rise from US$165.5 billion to about US$924.6 billion, lifting its share of the total market from 26.3% to about 55.9%. This reflects the combined effect of AI server investment, demand for HBM and high-capacity DRAM, and rising memory prices.

Alongside market growth, competition among nations in semiconductors is intensifying into a strategic rivalry.

The United States is a design powerhouse whose companies account for 53.4% of global semiconductor sales (2025). However, its share of global manufacturing capacity fell from 37% in 1990 to 10% in 2022 and, even with the effects of the CHIPS Act, is projected to reach only 14% by 2032. Washington is therefore pursuing the reshoring of advanced manufacturing alongside technology controls on China.

China is expanding its share of legacy logic capacity from 31% (2023) to 39% (2027, projected). It is rapidly closing the gap in NAND but still lags significantly in advanced DRAM and HBM.

Korea is overwhelmingly dominant in memory, holding 63% of DRAM revenue (Samsung Electronics 38%, SK hynix 25%) and 83% of HBM revenue in 2Q 2026, but it remains weak in foundry (Samsung Electronics 5.9%, 2Q 2026) and in its fabless ecosystem.

In Taiwan, TSMC holds 72.5% of the foundry market (2Q 2026), but its production is concentrated at home. Japan retains its edge in materials (56%) and equipment (32%), but its overall standing in the industry has weakened owing to a weak mass-production base in advanced logic. The EU has strengths in lithography equipment, anchored by ASML, and in automotive semiconductors, but depends on production outside the region for advanced logic and memory.

3. The State of China's Semiconductor Rise and Its Technology Level​

In terms of market share, signs of change are emerging in the established Big Three structure. In DRAM, CXMT (ChangXin Memory Technologies) is emerging as a meaningful fourth player alongside Samsung Electronics, SK hynix and Micron. CXMT's revenue share rose from 3% in 1Q 2025 to 8% in 1Q 2026 and 10% in 2Q 2026. Still, with a wide gap remaining to Micron (24%) in the same quarter, it is more appropriate to see this as an early stage of change in the Big Three structure. CXMT's growth also remains centered on commodity DRAM (DDR4, LPDDR, etc.) and the Chinese domestic market.

Beyond CXMT, new DRAM entrants are also appearing. SwaySure, backed by Shenzhen state capital, is building a 28nm-based 12-inch line (planned capacity of 140,000 wafers per month), and YMTC, whose main business is NAND, is pushing into DRAM, having asked customers to evaluate LPDDR samples.

In NAND, the market structure is changing faster than in DRAM. In 2Q 2026, revenue shares stood at Samsung Electronics 28%, SK hynix 19%, Micron 15%, Kioxia 14% and YMTC 14%, a market already reshaped into a multipolar structure.

In technology, China's catch-up is asymmetric, progressing in the order of NAND → commodity DRAM → HBM.

In NAND, YMTC is mass-producing 267-layer products and developing products with more than 300 layers, approaching the leading group. Compared with SK hynix's 321-layer products in mass production and Samsung Electronics' products of more than 400 layers being prepared for mass production, the gap is about one to two generations on a current-process basis. However, given the lack of data on yields and cost per bit for the latest products, real competitiveness cannot be judged by layer count alone.

In commodity DRAM, CXMT is mass-producing G4 (about 16nm class) and developing G5 (1a class, about 14-15nm class). Korean firms are already mass-producing 1b (about 12-13nm class) and transitioning to 1c (about 11-12nm class), putting the gap at about two to three generations by process, or about three to four years in terms of when China will reach the same process node.

In HBM, Korea's SK hynix began full-scale mass production of HBM3 in 2022 and HBM3E in 2024, and in 2026 Korean firms entered the stage of commercial HBM4 shipments. CXMT, by contrast, has both HBM3 and HBM3E still undergoing testing by Chinese AI chip companies and has yet to reach full-scale mass production. By product generation, China is about one to two generations behind Korea. HBM competitiveness, however, is determined less by whether a given generation has been developed than by stacking, packaging, mass-production stability and customer qualification. Taking these into account, China's real mass-production competitiveness is assessed to be about three to five years behind Korea's leading firms.

China's memory production capacity is expected to grow both in absolute terms and as a share of the major producers' combined capacity, while the share of Korean firms is expected to decline.

Measured by monthly wafer input, CXMT's DRAM capacity is projected to expand from 265,000 wafers (14%) in 2025 to 500,000 wafers (18%) in 2028. Over the same period, Samsung Electronics' capacity will rise from 705,000 (38%) to 880,000 (32%) and SK hynix's from 535,000 (29%) to 745,000 (27%), increasing in volume but declining in share. In NAND, YMTC will expand from 148,000 (11%) to 260,000 (16%), while the shares of Samsung Electronics, from 414,000 (31%) to 460,000 (29%), and SK hynix/Solidigm, from 233,000 (18%) to 275,000 (17%), are expected to decline.

4. Drivers of and Challenges to China's Catch-up​

Long-term funding and tax support from central and local governments are being layered into China's semiconductor industry. Between 2014 and 2020 alone, an estimated cumulative US$150 billion in public funds was injected through the central government's Big Fund and local government industrial funds. The Big Fund's registered capital has also grown about 3.5-fold, from RMB 98.72 billion in Phase I (2014) to RMB 204.15 billion in Phase II (2019) and RMB 344 billion in Phase III (2024). Manufacturers at 28nm and below that meet the requirements are exempt from corporate income tax for up to ten years.

More recently, capital markets have joined as a new source of funding. CXMT raised about RMB 57.9 billion (about US$8.6 billion) through its July 2026 listing on the Shanghai Stock Exchange's Science and Technology Innovation Board (STAR Market), and YMTC is also seeking to raise about RMB 33 billion (about US$4.9 billion). A virtuous funding cycle is taking shape: policy funds → corporate growth → IPO → reinvestment in capacity and R&D.

In terms of talent, China has a massive base, with about 12.22 million college graduates in 2025 and about 7.95 million full-time-equivalent person-years of R&D personnel. On top of this, it has institutionalized a sustained supply of master's- and doctoral-level talent through the designation of national model microelectronics schools (示范性微电子学院) in 2015, the 2016 Plan to Strengthen Integrated Circuit Talent Development, and the creation of Integrated Circuit Science and Engineering as a first-level academic discipline in 2020. It is also actively recruiting engineers with process, equipment and yield-management experience from Korea and Taiwan by combining national overseas talent programs with generous corporate compensation.

China relies on a governance system in which central ministries and local governments share roles under the strategic command of the Party Central Committee. The Central Science and Technology Commission, established in 2023, oversees national science and technology strategy, while an integrated circuit task force under the National Development and Reform Commission and local government task forces manage talent, policy, investment attraction and projects in an integrated manner.

China's vast domestic demand is another key driver. China is the world's largest production base, assembling about 35% of the world's electronic devices and accounting for about 30% of global manufacturing value added, and it simultaneously hosts large chip-consuming industries such as electric vehicles, smartphones, industrial robots and data centers. Because semiconductors are an industry in which quality and cost competitiveness accumulate through high-volume production and repeated process improvement, huge domestic demand provides the foundation for learning by doing: accumulated production experience → performance improvement → further adoption.

Policy-driven demand is creating early markets for domestic chips. Since 2025, the government has reportedly required that at least 50% of newly installed equipment be sourced domestically as a condition in the approval process for new or expanded fabs. Since 2026, government procurement has applied a preference scheme that deducts 20% from the evaluated bid price of products made in China.

Paradoxically, US sanctions are accelerating localization. US technology controls on China have expanded step by step, from Huawei's addition to the Entity List in 2019 to comprehensive controls in October 2022, the extension to software and HBM in December 2024, and outbound investment screening in January 2025. As the controls have become protracted, however, China has pursued both "design-out," which excludes US equipment, and "design-around," which circumvents controlled technologies, broadening domestic substitution. As a result, China's share of revenue at major foreign equipment makers has declined, while Chinese equipment makers are growing rapidly on the back of import-substitution demand.

China is also pursuing "overtaking on the curve" (弯道超车) through new technology pathways such as RISC-V and photonic chips. In 2023 it extended this detour into standards, establishing its own chiplet interface standard. This deserves attention because the rules of competition may change before the technology gap closes.

China's catch-up, however, also faces clear limits.

EUV is the product of a technology ecosystem that requires the combined accumulation of light sources, optics, materials, inspection, precision control and process integration, making it difficult to localize quickly. The EUV prototype China completed in early 2025 succeeded in generating light but has not yet produced chips. The government's target is 2028, but the industry sees around 2030 as a more realistic timeline. Workarounds such as DUV multi-patterning also cannot fully replace EUV because of the cost and complexity of additional process steps. 

US regulations also constrain Chinese firms' access to the US market and their ability to win global customers. Global PC makers' adoption of CXMT DRAM is likewise largely confined to non-US markets.

5. Implications of Korea's Semiconductor Industry​

China's supply expansion could squeeze the profitability of Korean firms. What Korea should watch is not when China overtakes Korea technologically, but when it establishes itself as a sustainable supplier in commodity markets. Counterpoint Research identifies a bit share of about 15% as the threshold for a late-entrant memory company's long-term survival and its ability to keep investing in next-generation technology. The evidence is the case of Taiwanese DRAM makers, which, after falling below this line in 2008, failed to secure funds for next-generation investment and were pushed into niche positions with shares of about 3%. In 2Q 2026, YMTC was already near the threshold, at about 14% of NAND by both revenue and bits. CXMT, by contrast, stood at 10% of DRAM revenue and about 9% of bits. Counterpoint Research's base scenario projects that CXMT's DRAM bit share will rise to about 11% by 2028, while its revenue share will actually fall to about 9%. This appears to reflect the normalization of commodity DRAM prices, whose recent surge has lifted CXMT's current revenue share, and the difference in product mix relative to incumbents with a larger share of high-value products such as HBM. The projected bit share is also lower than CXMT's projected capacity share (18% in 2028) because bits produced per wafer vary with yield and process node. In short, China has already reached the threshold of a sustainable supplier in NAND, while in DRAM the period around 2028 will show whether it crosses that line.

Backed by policy finance and local government support, Chinese firms can prioritize capacity and market-share expansion over short-term profitability, which could delay production cuts and investment adjustments even in downturns. In that case, price pressure on commodity DRAM and NAND could be prolonged, and the supply discipline centered on incumbent firms could weaken. Furthermore, if China becomes a persistent price competitor in commodity memory and begins supplying HBM reliably, the threat could extend to Korea's high-value memory profit base.

Korean firms could face a double squeeze: losing existing customers inside China while newly competing with Chinese firms outside it. In January-July 2026, 30.1% of Korea's memory chip exports went to China. As Chinese customers gain more domestic alternatives, Samsung Electronics' and SK hynix's bargaining power on price and supply weakens. Outside China, major PC makers such as HP, ASUS and Acer reportedly began in 2026 to adopt CXMT DRAM on a limited basis in some laptops for non-US markets. Apple has also reportedly considered sourcing CXMT DRAM and discussed approval with the US government, although whether it has actually done so has not been confirmed.

The growth of Chinese equipment makers also poses a challenge to Korea's materials, parts and equipment (MPE) industry. From January to May 2026, exports of semiconductor and display equipment totaled US$2.16 billion, down 0.4% year on year, revealing a gap between the chip export boom and the performance of the domestic equipment industry. This gap cannot be explained by China's equipment localization alone. However, with Naura rising from 8th in 2022 to 5th in 2025 in global equipment sales rankings and three Chinese firms, including Naura, AMEC and SMEE, entering the global top 20, the rapid expansion of import substitution in China, the world's largest equipment market, is a structural factor constraining the medium- to long-term growth of Korea's MPE industry.

Korean production bases in China are caught in a double squeeze between the United States and China. Samsung Electronics' Xi'an fab accounts for about 30-35% of the company's NAND production, while SK hynix's Wuxi fab accounts for about 35-40% of its DRAM production and its Dalian fab for about 35-40% of its NAND production (as of 2026), making them key sites that are hard to replace in the short term. The US revocation of VEU (Validated End-User) status and the CHIPS Act guardrails (which bar subsidy recipients from expanding advanced capacity in China by more than 5%) constrain equipment imports, capacity expansion and conversion to advanced processes. At the same time, firms that comply with US regulations risk countermeasures under China's Anti-Foreign Sanctions Law and its Regulations on Countering Improper Extraterritorial Jurisdiction by Foreign Countries, which took effect in 2026.

6. Policy Recommendations

Korea must continuously create strategic technology advantages beyond HBM. The current boom may mask structural change. How effectively Korea uses the time and profits it has now secured to build the next technology gap and industrial base will determine the future competitiveness of Korean semiconductors. Korea should strengthen its HBM4 mass-production competitiveness and advance customer qualification and the transition to mass production for HBM4E, while preemptively securing memory-compute convergence technologies such as advanced packaging, optical interconnects and PIM (processing-in-memory), as well as next-generation memories based on ferroelectric, magnetic and memristor technologies. The government should support long-term fundamental technologies, talent and research infrastructure, and build a stage-by-stage support system linking R&D → validation → mass production → customer qualification → standards and patents.

Korea must also strategically maintain its production and customer base in commodity markets. Commodity memory underpins investment in advanced technology through scale, process learning and cash generation. Korea therefore needs a cost structure, production flexibility and core customer base that can remain competitive even when prices fall. The government should continuously monitor China's capacity additions, prices, inventories and utilization rates, and apply policy finance, tax support and facility upgrade support flexibly. It should also nurture domestic demand industries such as AI, data centers, automobiles and robotics to secure stable anchor demand, and improve production conditions including power, water, land and permitting.

Semiconductor talent must be managed as a core national security asset. In semiconductors, the departure of a single key engineer can transfer tacit knowledge, such as process integration and equipment settings, accumulated over many years. Talent development, recruitment and retention must therefore be treated together with technology protection as a single national strategy. Korea needs flexible labor rules suited to the industry, job- and performance-based compensation, and world-class pay and conditions. Rather than relying solely on after-the-fact punishment, technology protection should be strengthened by raising incentives for long-term employment through compensation, career development and a better research environment, alongside protective measures at the point of retirement or job change.

MPE policy must be restructured so that it leads to actual adoption in mass production. Korea should continuously review foreign dependence and substitutability, focusing on critical items whose supply disruption would directly halt production, and raise supply-chain resilience through multiple suppliers, alternative technologies and strategic stockpiles. It should use the Yongin mass-production-linked mini-fab (total project cost of KRW 1 trillion, 2025-2031) to validate domestic MPE products in an environment close to actual mass production. In addition, it should share the risks of joint development between large chipmakers and MPE firms and of early application in mass production, and support evaluation and certification, so that validated products move through customer evaluation and qualification to adoption in mass production. It should also broaden the global customer base by supporting joint development, local testing and customer qualification with overseas customers.

Korea should strengthen technology alliances while preserving room for strategic cooperation with China. With the United States, Japan, the Netherlands and others, it should deepen cooperation in joint development, investment, production and procurement, focusing on next-generation technologies, key equipment and materials that Korea cannot secure on its own, and take the lead in international standards discussions in fields where rules have yet to be formed. The scope of cooperation with major partners should be adjusted to each country's industrial structure and interests, and firms' experience in investment, production and procurement should be reflected in policy and negotiations. While maintaining its technology alliance with the United States, Korea needs a national-interest-based response to US restrictions on investment in China, tariffs and demands for investment in the United States, one that protects its domestic production base and corporate competitiveness.

Cooperation with China should be differentiated by sector. Cooperation should be restricted in areas with a high risk of technology transfer, such as advanced memory, leading-edge processes, key manufacturing equipment and advanced packaging. By contrast, the operation of production bases in China and the local supply, joint testing and customer qualification of non-sensitive MPE and fabless products should continue. At the same time, Korea should actively consider mutually beneficial Korea-China cooperation in third countries where the two share interests, such as joint supply in non-sensitive areas, participation in local projects and mutual certification. For critical minerals and some materials and equipment, Korea should maintain trade with China while reducing excessive dependence through long-term contracts, strategic stockpiles and diversified sourcing.

Finally, a dedicated body within the government should continuously monitor US and Chinese regulations and changes in China's industry, technology and markets, and link government diplomacy with corporate diplomacy to minimize geopolitical risk while strategically leveraging market and investment opportunities.

 ​​