Richard Gray is an analyst who works for a foreign policy think tank in Washington, D.C.

Since 2025, South Korea’s capital markets have vacillated from the world’s best-performing to shedding more than $1 trillion in market capitalization. This volatility is a story of how Korea’s financial system overlays supply chain specialization. With Korean firms Samsung and SK hynix controlling a majority share of global memory, South Korea’s stock market functionally operates as a single, highly leveraged node within the AI financial architecture.

Two interwoven dynamics explain the convulsions of Korean capital markets and their global ramifications. The first is the Korea discount, in which Korean equities are structurally undervalued because of lagging financial development and weak corporate governance. The second is a memory discount, with each leading high-bandwidth memory (HBM) company (Micron in the U.S., SK hynix and Samsung in Korea) having depressed valuations in their respective markets.

The June-July 2026 crash rippled far beyond Korea’s borders and forewarns what may be to come in future market corrections. For Korea, its historically shallow capital markets and exposure to the memory sector — which is now facing emergent Chinese competitors — serve as overlapping pressures on its financial position. In this environment, the threat of a market correction cascading across global markets and the U.S.-China tech competition looms large.

The Korean market is essentially a bet on two companies competing in the same industry.

Korea’s memory chip industry has driven exponential year-on-year growth in Korea’s capital markets. From March 2025 to June 2026, the KOSPI — Korea’s main stock index — rose 241%, adding nearly $3.3 trillion in capitalization and exceeding the total increase accumulated over the index’s 43-year history (Bloomberg).

In June 2026, SK, Samsung, and Naver announced plans to build chipmaking plants and data center facilities worth $880 billion — a sum adding up to more than 50 percent of Korea’s current annualized GDP.

The Korean AI capital boom powered these commitments. In Q1 2026, SK hynix reported a quarter-on-quarter increase in revenue of 60%, while Samsung recorded 40% revenue growth.

This concentration bleeds into the global semiconductor supply chain. As of Q2 2026, SK hynix and Samsung collectively accounted for 83 percent of the global HBM market and 63 percent of the dynamic random-access memory (DRAM) market. Micron, the closest American competitor, maintains an 18 percent market share in HBM and 24 percent in DRAM.

This domestic and international centralization is one of the clearest financial expressions of the artificial intelligence boom. Alongside SK hynix and Samsung, Micron forms a memory chip triopoly. The emergence of CXMT — a firm with deep state backing from the world’s second-largest economy — introduces risks of China moving up another value chain, accelerating production, and drawing down market pricing. Such a conversion would challenge the position of the incumbent memory players, introducing new risks of shortfalls in valuation that could spread financial contagion across the AI stack.

Korea’s memory chip bet has since faltered. From June 22 to July 16, the KOSPI entered a bear market, shrinking more than a quarter and wiping out nearly $1.1 trillion in market capitalization. (CEIC) SK hynix and Samsung fell 38% and 29% over the same window. (Bloomberg)

The June-July bear market was as tied to an industry as a country. Micron’s shares fell approximately a third in the same span, while Chinese and Taiwanese semiconductor stocks also were being sold off.

As the Iran war’s reescalation put additional pressure on the Korean economy through currency pressures and rising energy costs, Korean regulators warned of chip-sector concentration as a market risk. As of July 2026, Samsung Electronics and SK hynix alone accounted for more than half of KOSPI’s total market capitalization. In contrast, the Magnificent Seven (seven highest-valued American stocks) taken together are 37.2% of the S&P500. (Morningstar)

Throughout the first half of 2026, foreign investments have flowed out of Asian equities at a record pace — with Korea among the hardest hit. In line with the KOSPI’s crash, in May and June, South Korea’s capital flight (in equities) exceeded $35 billion. On July 28, Korea’s stock market dropped more than 8% during intraday trading, the eighth instance that year. There have been at least 33 days of upticks and downturns of more than 5%.

In response to the market’s sharp decline, the Financial Services Commission (FSC) suspended sales of leveraged Exchange-Traded Funds (ETFs) tied to individual stocks “until market conditions stabilize,” and tripled the minimum deposit required to trade them to $20,300, effective August 5. While ETFs often are an investment vehicle that holds a broad set of assets, the ones popularized in the Korean market were tied to the performance of single stocks and acquired in high volume through loans. These ETFs let ordinary investors bet on a single company’s stock while magnifying gains and losses. After the ETFs launched in May, trading in SK hynix and Samsung and their ETFs came to dominate the entire Korean stock market, peaking at 84 percent of all volume in June.

This volatility is partially powered by South Korea’s shallow capital markets. Historically, Korea’s capital markets have been characterized by a structural undervaluation of listed enterprises known as the “Korea discount.” Financial analysis company MSCI continues to label Korea as an emerging market, citing limited convertibility of the won in the offshore currency market, excessive requirements for foreign securities, and limited investment instruments.

Korea’s comparatively low price-to-book ratio — a comparison of stock prices and a company’s net assets — introduces risks that accompany rapid capital market growth. Throughout 2025, Korea had the lowest price-to-book of any of the surveyed markets, well below Hong Kong, Malaysia, or Singapore.

Era Dabla-Norris of the International Monetary Fund argues that shallow markets with limited diversification lead to concentrated counterparty exposures, potentially increasing financial stability risks during periods of market deepening. As the KOSPI retains a globally competitive return on equity — how efficiently companies generate profits from shareholder equity — the “Korea discount” is likely driven by a combination of genuine mispricing and governance risk.

Memory has become the highest revenue-generating segment in Korea’s capital market and the global semiconductor supply chain alike. According to Gartner, memory revenue is projected to surpass $1 trillion and account for more than 55 percent of all semiconductor revenue by 2027.

The value of South Korea’s DRAM exports has increased more than fivefold year-on-year (CEIC). An increase in memory demand has led to Korea’s semiconductor export volumes more than doubling in that same period.

At the same time, there is a global memory shortage. Samsung has earmarked 70% of its memory production for long-term agreements through 2030. This supply-demand imbalance is expected to become more severe through 2028 (JP Morgan). SK hynix CEO Kwak Noh-Jung has stated that the memory shortage could persist through 2030. According to JP Morgan, DRAM prices rose more than 400% from 2024 to 2026, with greater demand and long-term arrangements increasing overall memory pricing and reducing available supply.

Historically, memory has been classified as cyclical, requiring intensive capital investments. Competition was primarily on the lines of affordability as opposed to quality, with an ebb and flow in demand and downturns shrinking the number of players within the market. There is an ongoing debate on the extent to which the AI boom has reshaped the memory industry away from its cyclical nature. Unlike in the past, the market concentration among three incumbents — and CXMT as an emergent player — with the specialization and value-add of HBMs is thought to have transformed into a “differentiated product.”

Such a proposition relies on a continued and relatively stable demand to fuel the scale of ongoing investments. While long-term contracts provide immediate stability, they will also tie the memory industry’s future to the ability to extend them. Korea’s recent memory chip manufacturing investments are intended to be operational in 2030, just as Kwak Noh-Jung predicts that the memory chip shortage may come to a close. What then?

Maintaining the balance sheets and capital access of these national champions, therefore, is a much more expansive initiative than Korean financial stability alone.

With CXMT looking to enter the HBM market, demand is not the only factor at play. On August 31, 2026, The Information reported that CXMT met its objective of producing HBM3E — the fifth-generation HBM architecture — with the intention of expanding production in 2027. This situates CXMT two generations behind SK hynix and Samsung, which have begun shipping samples of their seventh-generation HBM4E, and one generation behind Micron’s HBM4. What is unclear is Samsung and SK hynix’s timeline for mass production of HBM4E, as well as the extent to which CXMT will be able to leapfrog generations.

CXMT has become a growing DRAM player and emergent HBM challenger. According to Reuters, Hefei “government-linked investors” and Anhui’s “provincial government-backed entities’ stake” account for roughly half of CXMT’s ownership, with industrial subsidies and state support contributing to the firm’s rise as China’s DRAM national champion. CXMT has continued to scale its market share in DRAM, growing from 3 percent in Q1 2025 to 10 percent in Q2 2026. In its first earnings report, CXMT reported that its H1 2026 profit was $22.4 billion after sales increased by 874%. Its market cap, as of August 28, stood at 4.01 trillion Chinese yuan.

In December 2024, the Biden administration restricted the export of DRAM and HBM to Chinese end users. In less than two years, CXMT has gone from prototyping the earliest generations of HBM to only two generations behind the cutting edge. This foothold within China — which remains its immediate priority, with HBM3E shipments to Alibaba and Cambricon Technologies — has facilitated its growing share of the global market. However, as Apple considers using CXMT as a memory supplier — even as the Trump administration opposes such sourcing — the existing shortages and price hikes could position Chinese suppliers to sell to American device and data center firms. This asymmetry of CXMT’s potential access to Chinese and global markets (assuming it is not added to Commerce’s Entity List), while Micron, SK hynix, and Samsung remain blocked from the Chinese market, could eventually compress margins and increase competition.

This market landscape and historic cyclicality materialize as a “memory discount”, where memory stocks have lower forward price-earnings ratios (PE) — current share price compared with expected earnings — than other equities on their listed exchange. This is true for SK hynix and Samsung, which are lower than the surveyed Korean firms Naver, Hanwha Ocean, and Kakao, as well as for Micron, which has a similar PE ratio to SK hynix and Samsung despite the U.S. having significantly deeper capital markets.

Of the memory players, the only current exception to this is CXMT, which recently listed on July 27. CXMT, in fact, outperforms Baidu, Alibaba, Tencent, and Xiaomi on a PE basis. As equities are one lever for companies to access capital, CXMT’s unique status as China’s only near-frontier player in a securitized, high-profit industry could provide it with unique, high levels of state support, reflecting in its valuation.

When I previously posed a related question on the nature of China’s capital markets to Zongyuan Zoe Liu, Senior Fellow at the Council on Foreign Relations, she said that: “Importantly, the Chinese government has been encouraging long-term patient capital to be deployed in China’s equity market [for national champions]. One such example is the encouragement of insurance companies and pension funds to invest in the equity market.” CXMT’s market capitalization, therefore, is as much a financial position as national industrial policy.

While successive Korean governments have attempted to close the “Korea discount”, this sectoral discount complicates the overall path to capital market deepening. Interestingly, however, the KOSPI bear market had an effect across sectors, with downturns in shipbuilding, chemicals, construction, and steel (CEIC). So, while the upside is contingent on memory performance, the downturn is cross-sectoral.

For Korea, there are market, political, and fiscal incentives to deepen capital markets. Deepening capital markets is viewed to diminish price disruptions, which is amplified by Korea’s sectoral concentration. Improving Korea’s investment environment is fundamentally a political question, both in terms of credibility, enforcing government policies across political cycles, and deterring cross-shareholding structures that entrench controlling shareholders. On the fiscal front, the National Pension Service holds roughly 7 percent of Korea’s stock market capitalization, so any re-rating of Korean equities directly revalues the fund’s existing holdings. As Korean society ages, higher returns are necessary to ensure adequate future payouts.

In response to a perceived valuation gap between Korean and international equities, successive Korean governments have increased access to financial instruments and furthered corporate governance reform. Since President Lee took office, the following measures have been introduced or implemented:

• Financial Instruments: In July 2026, Korea introduced 24-hour foreign exchange trading, with an offshore settlement system intended to enable registered foreign institutions to trade and source the won freely. Additional reforms include lifting the prohibition on short selling and easing foreign investor access by replacing the Investor Registration Certificate with Legal Entity Identifiers. MSCI notes it is tracking implementation, with the won not deliverable offshore and onshore liquidity during extended FX hours lagging developed markets.

• Corporate Governance: The Commercial Act was amended in 2025 and 2026, now requiring directors to have fiduciary responsibility to all shareholders, prohibiting companies from dual-listing subsidiaries, and restricting the number of issued treasury shares and their entitlement to voting and dividend entitlements. Combined with dividend tax reductions, reducing the maximum rate from 45 to 30 percent, these measures aim to increase capital efficiency and strengthen minority shareholder protections, reducing risk while improving performance.

South Korea’s memory champions are a single node in an AI financial architecture. According to Gartner, in 2026 companies will spend $2.6 trillion on AI (and nearly $3.5 trillion in 2027), more than 50 percent of which is for “AI Infrastructure.”

With single-stock ETFs, memory scarcity, and concentrated exposure, Korea’s volatility has amplified to global effect. At the height of the KOSPI crash, Evercore wrote that Korea had become the “‘Tail’ that ‘Wags the Dog’ of global asset markets broadly”, with a correlation of “an astounding .95” between the KOSPI and Nasdaq 100. During this period, Leopold Aschenbrenner’s Situational Awareness lost 67%, with leveraged investments resulting in liquidation upon a market correction.

The potential of AI as a technology has little relevance to the potential of an intermediate financial shock. Technology is not a unidirectional process of exponential growth, but of ups and downs where investments can underperform in the short-term and overperform in the long-term.

The growth in capital expenditure does increase the overall risk of misallocation and valuation fluctuations if, at least in the short term, returns on investment do not materialize. As Washington Crossing Advisors writes: “The likely long-term winners will be companies that can run a marathon, not a sprint… They must fund innovation… and they must get through the cooling-off periods.” Firms that can absorb a market correction — or that receive sufficient government backing — therefore, may be positioned to capitalize on a downturn.

Korea’s industrial investments reflect a commitment to maintain its advantages in memory. The $135 billion National Growth Fund is a sector-focused industrial policy initiative, of which artificial intelligence and semiconductors are two of the twelve target industries. K-Moonshot, similarly, is a $7.5 billion R&D platform, funding twelve projects to “aggressively integrat[e] artificial intelligence.”

While these initiatives could strengthen Korea’s competitiveness and drive economic growth, how and where resources are allocated should be done carefully, with deterring market swings as the principal objective.

Moving forward, Korea requires sustainable equity growth by hedging against risk — with the understanding that these are marginal efforts that may not overcome global financial trends. This can be pursued through two joint efforts of diversification.

The first effort is constraining access to leveraged investment instruments. Korea’s stock market is a de facto semiconductor ETF, and the trading of single-company, two-company, or single-industry ETFs should be banned indefinitely.

The second effort is broadening the market share of alternate sectors. While the June-July bear market indicates that downturns affected a broad set of sectors, greater KOSPI diversification would reduce the contagion effect of memory upswings and downturns.

As one example, Korea’s defense sector may diminish sectoral reliance. According to Dina Ting of Franklin Templeton, Korea’s dual exposure to AI and defense “helps explain why South Korea’s recent rally may be more durable than in past cycles.” From 2021-25, Korea accounted for 3 percent of global arms sales, a 24 percent increase relative to 2016-20. South Korea’s growth in defense arms sales covered Middle Eastern, European, and Indo-Pacific theaters, representing 47 percent of Poland’s weapons imports, 10 percent of the UAE’s, and 42 percent of the Philippines’.

This defense sector growth is likely to continue. At the Seoul International Aerospace and Defense Exhibition 2025, President Lee stated that it was his goal for Korea to become “one of the top four powerhouses in the defense industry.” As the Korean defense industry develops, the Financial Services Commission and the Ministry of National Defense should coordinate on a listing incentive program to bring emerging medium-sized defense enterprises to market, increasing the market share of non-chip sector companies over time.

As Korea capitalizes on strategic industries, prioritizing these emergent spaces — including, but not limited to defense, chemicals, robotics, etc. — to dilute any sectoral impact is crucial.

Historical precedents are pervasive in Korean financial thinking. Discussing the proposed $350 billion package of Korean investment in the United States, President Lee warned that without a currency swap, “South Korea would face a situation as it had in the 1997 financial crisis.” More recently, he turned the same historical lens on the domestic property market, cautioning that Korea may be “racing toward” the kind of real estate “bubble” that preceded Japan’s “lost 30 years.” Once again, Korea sits at the center of transnational financial instability.

Looking ahead, McKinsey projects that support for AI-related data center capacity could require upwards of $8 trillion by 2030, of which memory chips are a key pillar.

Korea’s stock market is simultaneously shallow (thereby volatile) and concentrated (thereby vulnerable to market cycles and emergent competitors). With Korea’s position in the AI supply chain and the scale of capex, these questions are worth considering:

1. What is to happen to these investments if another, more than $1 trillion shock hit the Korean market? What about the asset management firms that hold Korean equities — and those that bought in the preceding bear market?

2. While SK and Samsung have memory shortages and backfilled orders through the end of the decade, they are continuing to expand capacity. What would happen to their balance sheets if memory re-commoditized, CXMT became a strong HBM competitor, and prices collapsed?

3. As it begins its HBM drive, what would expanded state support of CXMT mean? If Micron, SK, and Samsung are sold for a “memory discount”, can Chinese industrial policy induce a premium to the national champion in a manner that neither the United States nor Korea could replicate?

4. Conversely, what would the ramifications be for the financial positioning of Micron, SK hynix, and Samsung if CXMT continues to gain market share at competitive prices in legacy DRAM, even if they don’t advance to the frontier of HBM?

5. American frontier labs have high valuation and low revenue; Chinese firms have suppressed valuations due to state support; and Korean memory producers have exponentially growing revenue yet carry a discount. What, if anything, would cause these dynamics to flip?

The technological and financial scale of these questions — and the people, countries, and enterprises embedded within them — is enormous. While these questions are not immediately answerable, it is vital for decision-makers to collectively confront their implications, assess the risk factors, and address existing vulnerabilities. The resilience of an allied AI stack, global financial stability, and prosperous economies all depend on it.

There is also a human story to this volatility. For The Wall Street Journal, Jiyoung Sohn, Sooyoung Rhee, and Jack Pitcher provided a snapshot into the “victims” of the KOSPI downturn. This is people’s livelihoods evaporating in an instant, with overleveraged, single-company and single-industry ETFs being legalized and encouraged at a time when Korea’s capital markets had not deepened. Such a system was poorly conceived, with the margin-called retail investors — from teachers to accountants to software engineers — unable to play the long game weathered by financial institutions.

As AI investments continue to climb globally, a downturn “loser base” will expand from retail investors to multinationals, nation-states, and institutional investors. These stories — and their long-term, macro ramifications — are worthy of remembrance.