South Korea AI Stocks: Samsung, SK Hynix and Leverage Risk

South Korea’s equity market has become the clearest example yet of a difficult truth about the artificial-intelligence investment cycle: exceptionally strong corporate earnings can coexist with exceptionally unstable stock prices. Samsung Electronics and SK Hynix are reporting record or near-record results as demand for high-bandwidth memory, server DRAM and enterprise solid-state drives accelerates. Yet the same two companies helped pull the benchmark KOSPI through one of the most violent boom-and-bust episodes in its history after leveraged single-stock products, retail borrowing, foreign portfolio rebalancing and concentrated index weights interacted with one another.

The immediate answer for investors is therefore more nuanced than either “Korean fundamentals are strong” or “the Korean market was a bubble.” The underlying businesses, especially the memory-chip leaders, are producing extraordinary cash flows and have secured multi-year demand commitments from major technology customers. The market structure around those businesses, however, became dangerously reflexive. Instruments designed to deliver two times the daily move in Samsung Electronics or SK Hynix were forced to add exposure as the shares rose and reduce it as they fell. Because the two stocks together represented more than half of the KOSPI at points in 2026, those mechanical flows could move not only the stocks but the entire national benchmark.

South Korean authorities have responded with a temporary halt on new single-stock leveraged product listings, restrictions on marketing, tougher education requirements, tighter controls on premiums and discounts, and a higher cash-deposit threshold. Additional measures under discussion include limiting such products to a portion of an investor’s portfolio and raising trading friction. Early data indicate that trading volume in the most popular leveraged products has dropped sharply. That may reduce the most destabilizing feedback loops, but it does not remove the risk available through overseas listings in Hong Kong and the United States, nor does it resolve the structural concentration of the Korean market.

The central investment question is no longer whether AI needs memory. It does. The more difficult questions are how long the current shortage and pricing power can last, whether hyperscaler capital expenditure can remain economically productive, how quickly Samsung and SK Hynix will expand supply, whether Chinese competitors can narrow the technology gap, and how much of the expected earnings through 2027, 2028 and beyond is already reflected in valuations. These questions matter to U.S. investors because Korean memory companies sit inside the same AI infrastructure chain as Nvidia, Microsoft, Alphabet, Amazon, Meta Platforms, Oracle, Broadcom and Micron Technology.

Last updated: August 4, 2026, 12:00 p.m. Eastern Time. Market figures and fund holdings may change after this research cutoff.

Key Takeaways

• The core distinction: Korea’s chip-company earnings remain powerful, but the equity market experienced a leverage and liquidity shock that was not explained by business fundamentals alone.
• Regulatory response: South Korea halted new single-stock leveraged product listings on July 16, 2026, banned marketing, strengthened education and premium-discount controls, and accelerated a KRW30 million cash-deposit requirement to July 31.
• Market concentration: Samsung Electronics and SK Hynix together represented slightly more than half of the KOSPI during the rally, making the national index unusually sensitive to two AI-memory stocks.
• SK Hynix: The company reported second-quarter 2026 revenue of KRW79.3 trillion and operating profit of KRW60.5 trillion, while emphasizing HBM, AI-server memory and long-term customer agreements.
• Samsung Electronics: Second-quarter 2026 consolidated revenue reached KRW171.5 trillion and operating profit KRW89.5 trillion, with the Device Solutions division generating almost all group operating profit.
• The valuation debate: Rapid earnings revisions can make headline price-to-earnings ratios look inexpensive even after a major rally, but the ratios depend on unusually high margins and optimistic forward estimates.
• Beyond semiconductors: Korean banks, insurers, holding companies, electrical-equipment makers, shipbuilders, defense contractors and auto suppliers offer exposure to the broader industrial and governance reform story.
• What matters next: Investors should watch leveraged-product assets, foreign flows, HBM contract terms, capital expenditure, shareholder-return policies, hyperscaler cash generation and evidence of supply discipline.

The Real Story: Strong Companies Inside an Unstable Market Structure

Sean Taylor, chief investment officer and portfolio manager at Matthews, framed the Korean opportunity around a distinction that is easy to lose during a market panic. Leverage may have added fuel to the rally and the correction, but the underlying demand for memory used in AI systems is not imaginary. Samsung Electronics and SK Hynix occupy critical positions in DRAM, NAND flash and high-bandwidth memory, or HBM. Their products are needed to move data rapidly between processors and memory in AI servers, and supply has struggled to keep pace with demand.

That argument is supported by the companies’ financial statements. SK Hynix’s second-quarter revenue increased 257% from a year earlier, while operating profit rose 557%. Samsung Electronics’ Device Solutions division reported KRW127.5 trillion in revenue and KRW89.2 trillion in operating profit during the same quarter. Both companies described robust server demand, constrained supply and growing interest in multi-year agreements.

Yet Taylor also described a market that had become too crowded. The concentration was not only conceptual, with investors worldwide chasing the same AI theme. It was mechanical. Domestic single-stock leveraged products launched on May 27, 2026, after rule changes allowed products with exposure of up to 200% to qualifying blue-chip shares. Similar products had already been available in Hong Kong, and U.S.-listed leveraged vehicles added another source of demand and forced rebalancing. Assets linked to Samsung Electronics and SK Hynix rose rapidly, then collapsed as the underlying shares reversed.

For a conventional long-only investor, a share price decline creates a choice: hold, buy, or sell. A daily leveraged ETF does not have the same freedom. It must reset its exposure to deliver its stated multiple for the next trading day. If its underlying stock rises, the fund generally needs to increase exposure. If the stock falls, the fund generally needs to reduce exposure. That procyclical behavior can reinforce momentum in both directions. In a market where two stocks dominate index weight and trading activity, the effect can spread far beyond the products themselves.

The July selloff demonstrated that a sound fundamental thesis can become a dangerous trade when too much capital reaches for the same exposure through instruments that must rebalance. The distinction matters because it affects how the episode should be interpreted. A leverage unwind does not prove that AI memory demand is about to collapse. Record earnings do not prove that the shares were correctly priced at their peak. Both statements can be true: the companies can be structurally important and the trade can become technically unsustainable.

A Timeline of the Korean AI Rally and Leverage Unwind

April 21–28: The rules change

On April 21, 2026, South Korea’s Financial Services Commission announced that revisions to the Enforcement Decree of the Financial Investment Services and Capital Markets Act would take effect on April 28. The old framework generally required an ETF to hold at least ten underlying securities and limited the weight of an individual constituent. The revision created an exception for qualifying blue-chip stocks and allowed single-stock products with exposure up to 200%.

The policy was intended to close a regulatory gap. Korean investors were already able to trade overseas-listed products tied to Korean companies, including products in Hong Kong. Regulators argued that allowing domestic products could keep activity inside a framework with stronger investor protection. The original safeguards included additional education, quizzes, clearer product labeling and a KRW10 million deposit requirement.

That logic was not unreasonable. Prohibiting domestic products while investors used less controlled foreign alternatives could simply export the risk. The problem was that the launch occurred while AI-memory shares were already attracting enormous speculative interest. The domestic products did not enter a calm market; they entered one of the most crowded momentum trades in the world.

May 27: Domestic single-stock leveraged products begin trading

The first Korean products launched on May 27. Demand concentrated quickly in vehicles linked to Samsung Electronics and SK Hynix. The products promised a magnified daily return, not twice the long-term return. That distinction is central. Investors who treated them as buy-and-hold substitutes for ordinary shares were exposed to path dependency, financing costs, derivative spreads, rebalancing effects and volatility drag.

June: Assets, share prices and index concentration surge

By late June, assets in Korean-underlying leveraged products had expanded dramatically across domestic and overseas markets. The KOSPI had nearly doubled during the first half of 2026, while Samsung Electronics and SK Hynix became so large that foreign funds were forced to reconsider position limits and benchmark exposure. Reuters reported that overseas investors sold a net $70.8 billion of South Korean equities in the first half, with much of the flow attributed to rebalancing and profit-taking rather than a rejection of the companies’ earnings outlook.

This is a counterintuitive but important point. A stock can be sold because it became too successful. When a company’s index weight rises rapidly, an active fund that does nothing can become more concentrated than its risk policy allows. A benchmark-aware manager may sell simply to avoid allowing one country, sector or company to dominate the portfolio. The same dynamic can occur in passive and rules-based products when index caps or diversification requirements are reached.

July 14–16: Volatility exposes the feedback loop

By mid-July, the KOSPI’s volatility index had reached levels far above its historical range. Samsung Electronics and SK Hynix together accounted for slightly more than half of the benchmark. Retail investors had become the primary buyers as foreign institutions reduced exposure. Margin financing on KOSPI shares approached record levels.

On July 16, regulators announced the first major package of restrictions. New listings were suspended, marketing was stopped, the planned deposit requirement was increased and investor education was strengthened. The Financial Services Commission also sought tighter management of the difference between a product’s market price and the value of its underlying holdings, an issue that becomes especially important when liquidity is poor or demand is one-sided.

July 24–31: Rules are accelerated as the selloff intensifies

Authorities originally expected the KRW30 million deposit requirement to take effect in early August, with the shift to cash-only recognition following later. They accelerated the rule to July 31. From that date, investors needed at least KRW30 million in cash in their accounts to make new or additional purchases of domestic or overseas single-stock leveraged products. Securities firms could no longer relax the threshold after investors accumulated experience.

The market did not stabilize immediately. On July 28, the KOSPI fell nearly 11%. On July 29, it dropped as much as 12.6% before closing down 6%, and a trading halt was triggered. SK Hynix fell 9.6% after record profit missed elevated expectations, while Samsung Electronics declined 5.2%. The two-day collapse erased an extraordinary amount of market value and produced forced selling among leveraged retail accounts and funds.

The next reversal was nearly as violent. The KOSPI surged 17.9% on July 31 as foreign investors bought a record KRW7.2 trillion of shares. That rebound did not prove that the market had found a durable bottom. It did show that after leverage is removed, a market with limited liquidity and heavy short positioning can move sharply upward when buyers return.

August 3–4: Volumes fall, but policy work continues

By August 3, assets in leveraged products tied to Samsung Electronics and SK Hynix had fallen from roughly $50 billion in late June to about $17 billion, according to research cited by Reuters. Trading in the leading domestic leveraged products also declined. On August 4, Reuters reported that daily volume in the KODEX SK Hynix single-stock leveraged ETF had dropped to KRW560 billion from KRW1.3 trillion on the previous Friday. Volume in the Samsung-linked product fell to KRW234 billion from KRW511 billion.

Trading in such products represented 33.4% of KOSPI market volume on July 30, then fell to 6.6% on July 31 and 5.4% on August 1, according to an analyst cited by Reuters. The decline suggests that the new barriers and the losses themselves are reducing activity. Authorities nevertheless continued to discuss additional restrictions, including a potential cap limiting single-stock leveraged products to 20% of an individual’s total investment assets.

How a Daily Leveraged ETF Can Amplify Both Gains and Losses

A two-times daily ETF seeks to deliver approximately twice the percentage return of an underlying asset for one trading day, before fees and other costs. It does not promise twice the return over a week, a month or a year. The difference arises because the fund resets its exposure every day.

Consider a stock that starts at 100, falls 10% to 90, then rises 11.11% back to 100. The stock finishes unchanged. A two-times daily product would fall roughly 20% on the first day, from 100 to 80. On the second day, it would gain about 22.22%, taking it to approximately 97.78. The underlying stock recovered fully, but the leveraged product remained down about 2.22%, before expenses.

The problem becomes more severe as volatility increases. If a stock falls 20% and then rises 25%, it also returns to its starting level. A two-times daily product would fall roughly 40%, from 100 to 60, then gain 50%, reaching 90. The stock is flat; the leveraged product is down 10%. This is often described as volatility drag or negative compounding, though the precise result depends on the path of returns.

Trending markets can produce the opposite effect. If a stock rises on consecutive days, daily compounding can generate more than twice the cumulative return. That feature attracts momentum traders and explains why the products can appear extraordinarily successful during a steady rally. The same mechanism becomes punishing when the direction reverses or daily swings increase.

The market impact extends beyond the investor’s account. To maintain two-times exposure, a fund generally uses swaps, futures or other derivatives and adjusts its position as the underlying asset moves. When the stock rises, the fund may need additional exposure near the close. When it falls, the fund may need to reduce exposure. Those trades can reinforce the day’s direction, particularly when the product is large relative to the liquidity of the underlying stock.

Single-stock leveraged products are more concentrated than broad-index leveraged ETFs. A broad index includes companies from multiple industries and may absorb company-specific shocks. A product tied to one stock has no such diversification. If the company reports disappointing earnings, faces a regulatory problem, loses a customer or becomes part of a sector-wide selloff, the leveraged product can lose a large portion of its value in one session.

South Korea’s case was unusual because the underlying stocks were not merely large companies. They were the dominant components of the national benchmark. That meant rebalancing in leveraged products could affect the shares, the shares could affect the KOSPI, the KOSPI move could trigger additional risk controls and margin calls, and those actions could feed back into the same stocks. This is the type of loop that can turn an ordinary correction into a market-structure event.

Why South Korea Was Especially Vulnerable

Two stocks dominated the benchmark

Samsung Electronics and SK Hynix together accounted for slightly more than half of the KOSPI during the rally. In the MSCI Korea Index at the end of July 2026, Samsung Electronics represented 34.39%, SK Hynix 32.23% and Samsung preferred shares another 3.75%. The exact weights differ by index methodology, free-float adjustments and rebalancing date, but the conclusion is consistent: Korean large-cap exposure became overwhelmingly tied to two memory companies.

Concentration produces both power and fragility. When the dominant companies report accelerating profit, the entire market can outperform. When investors reduce exposure to those companies, diversification elsewhere in the index cannot absorb the move. Banks, automakers, defense companies and shipbuilders may be performing well, but they cannot offset a double-digit decline in companies that represent half the benchmark.

Retail participation replaced foreign demand

Foreign investors sold heavily during the first half of 2026. Reuters reported $70.8 billion of net outflows from South Korean shares during that period and nearly $110 billion by mid-July. Much of the selling reflected portfolio rebalancing after Korean equities rose so rapidly. Domestic retail investors bought into the gap, purchasing KRW42.4 trillion of KOSPI shares in June and another KRW13.2 trillion during the first half of July.

Retail participation is not inherently destabilizing. Individual investors can provide liquidity, identify mispriced companies and support long-term ownership. The risk arose from the combination of borrowed money, concentrated positions and leveraged products. Margin financing on KOSPI shares stood at KRW28 trillion on July 14, close to the June 24 record of KRW29.8 trillion. When prices fell, brokers reduced exposure and forced sales at the same time leveraged funds were rebalancing downward.

Overseas products limited domestic regulatory control

South Korean authorities could restrict new domestic listings and impose cash-deposit requirements on Korean brokerage customers, but leveraged products tied to Samsung Electronics and SK Hynix also traded in Hong Kong and the United States. Regulators recognized this cross-border problem by applying the deposit requirement to both domestically and overseas listed products purchased through Korean securities firms. Even so, product design and market-making standards outside Korea remained under foreign regulators.

Liquidity was uneven

A company can have a huge market capitalization without offering unlimited trading liquidity. Large blocks, strategic ownership, preferred shares and the behavior of long-term holders can reduce the free float available to absorb rapid flows. During stress, bid-ask spreads widen, market makers become more cautious and derivative hedges become more expensive. The result is that the price impact of a given order can increase just as funds need to trade most aggressively.

The AI trade had become globally crowded

Korea’s rally was part of a broader concentration in AI infrastructure. U.S. hyperscalers, semiconductor designers, foundries, networking companies, power-equipment manufacturers and data-center operators all attracted capital. Investors increasingly owned similar exposures across regions: Nvidia in the United States, TSMC in Taiwan, Samsung Electronics and SK Hynix in Korea, and memory or equipment suppliers throughout Asia. A change in confidence about AI capital expenditure could therefore trigger simultaneous selling across markets.

What the New Rules Can Fix—and What They Cannot

The first purpose of the Korean measures is to slow the entry of new speculative capital. A KRW30 million cash requirement raises the threshold for participation and prevents investors from using the market value of other securities as a substitute for cash. Recognizing proceeds only after settlement further reduces the ability to sell one asset and immediately redeploy unsettled funds into a leveraged product.

The marketing ban addresses another concern. Promotional campaigns can make complex products appear simple, particularly when recent returns are spectacular. A product that doubled during a steady rally can be advertised through performance charts without making the path-dependent risks intuitive. Stopping promotional events may reduce momentum-driven account openings and impulse purchases.

Enhanced education is also relevant, but its effectiveness depends on whether investors understand rather than merely complete the requirement. The revised framework adds a third hour of instruction, more extensive questionnaires and a passing score. The curriculum includes actual loss cases, negative compounding and premium-discount behavior. Those are the correct topics. The unresolved question is whether a short course can overcome the emotional appeal of a rapidly rising market.

Tighter management of premiums and discounts can protect investors from purchasing a product far above the value of its holdings. During demand surges, an ETF can trade at a premium to net asset value if creation activity, hedging capacity or market making cannot keep up. During a panic, the reverse can occur. Strengthening liquidity-provider obligations from a 3% to a 2% management threshold may reduce extreme dislocations, although it cannot guarantee that market price and net asset value remain aligned in a disorderly market.

The proposed increase in the minimum trading lot from one share to 20 shares would raise the practical cost of entering a domestic single-stock leveraged position. It is a blunt tool. Larger lots can reduce small speculative trades, but they also reduce flexibility and may concentrate activity among wealthier traders. A portfolio-level cap, such as the 20% limit discussed by authorities, would target concentration more directly.

None of these measures changes the daily reset mechanism. The products will still tend to buy after gains and sell after losses. Nor do the measures reduce the KOSPI weight of Samsung Electronics and SK Hynix. Index concentration will remain as long as the companies’ market values dominate Korea’s listed universe. The rules can reduce the size of the accelerant; they cannot remove the underlying structure.

The rules also create the possibility of migration. Investors blocked from domestic products may use foreign listings, options, futures, margin loans or structured notes. Regulation of one instrument can shift risk rather than eliminate it. Authorities therefore need to monitor total economic exposure, not merely assets in domestic ETFs.

A durable solution will require better product suitability standards, transparent leverage data across markets, coordination with foreign regulators, stronger margin controls and a broader Korean equity market in which more companies grow large enough to reduce benchmark concentration. Corporate governance reform and capital allocation will matter as much as product regulation because they can help attract long-term capital into a wider range of businesses.

SK Hynix: Record Operations, Elevated Expectations

SK Hynix entered 2026 as one of the clearest corporate winners from the AI infrastructure boom. Its lead in high-bandwidth memory gave it a critical role in systems built around Nvidia accelerators and other high-performance processors. HBM is not simply additional memory capacity. It is designed to deliver very high bandwidth with lower power consumption than conventional architectures, an essential requirement for training and running large AI models.

For the second quarter of 2026, SK Hynix reported revenue of KRW79.3187 trillion, operating profit of KRW60.5426 trillion and net profit of KRW93.9226 trillion. Revenue rose 257% from the second quarter of 2025, while operating profit increased 557%. The operating margin reached 76%. Those figures are extraordinary by any historical semiconductor standard.

The net profit figure requires context. It exceeded operating profit because of non-operating factors, including gains associated with the company’s stake in Kioxia, according to analyst commentary reported by Reuters. Investors should not treat a one-time or valuation-related gain as equivalent to recurring operating earnings. The operating result is the cleaner measure of the core memory business, though even that result reflects unusually strong pricing and supply conditions.

SK Hynix said both DRAM and NAND prices increased significantly from the previous quarter. High-value products, including HBM, AI-server DRAM and enterprise SSDs, improved the sales mix. Cash and cash equivalents reached KRW88 trillion, while total debt declined to KRW18.6 trillion, producing a KRW69.4 trillion net cash position.

The balance-sheet transformation is strategically important. Memory manufacturing is capital intensive and historically cyclical. Companies often invest aggressively during booms, create too much capacity, then suffer when prices collapse. A larger cash position allows SK Hynix to fund fabrication plants, advanced packaging and research without relying excessively on debt during the next downturn.

Long-term contracts reduce volatility but may cap upside

SK Hynix said it had concluded discussions on approximately ten long-term supply agreements and was negotiating with additional customers. Typical agreements can last five years and include deposits or other safeguards. This changes the economics of memory in several ways.

First, contracts provide demand visibility. A manufacturer can commit capital to new capacity with more confidence if customers guarantee purchases or provide upfront payments. Second, contracts can reduce the risk of a sudden inventory correction. Third, they can strengthen customer relationships around customized products, packaging and future HBM generations.

The tradeoff is that contracts may limit participation in future spot-price spikes. If the market becomes even tighter, a supplier with a large volume committed at negotiated terms may earn less than one selling more capacity at market prices. Investors reacted to that possibility after the earnings release. The company’s operating profit, though record-high, missed elevated analyst expectations, partly because HBM pricing increased less than conventional DRAM and because some advanced-product shipments were delayed.

This is a classic example of the difference between business quality and earnings surprise. SK Hynix’s business was not weak. The company generated record operating profit and described demand as robust. The stock fell because expectations had moved even faster than the results. In a market priced for repeated upside surprises, an excellent quarter can be disappointing.

Capital expenditure is the key discipline test

SK Hynix planned to increase 2026 capital expenditure to the high-KRW40 trillion range, up from KRW30.2 trillion in 2025. The company argued that it would align investment with confirmed demand and avoid creating excess supply. That promise deserves close attention.

HBM capacity requires more than ordinary wafer output. Advanced stacking, packaging, testing and access to specialized equipment can create bottlenecks. Expanding one part of the chain without the others may not translate into saleable product. Conversely, coordinated expansion across DRAM wafers, packaging and customer qualification can release large amounts of supply. The timing of that capacity determines whether current pricing power lasts.

Investors should therefore follow capital expenditure not only as a total figure but by category: leading-edge DRAM, HBM packaging, NAND, enterprise SSDs and infrastructure. They should also compare investment with customer prepayments and contracted volumes. Capital backed by enforceable demand commitments is less speculative than capacity built around optimistic forecasts.

Shareholder returns remain unresolved

Record cash generation created pressure for a clearer shareholder-return policy. SK Hynix said it planned to provide more detail later in 2026 but had not finalized timing, size or structure at the second-quarter announcement. The company must balance dividends and repurchases against the need to fund a capital-intensive technology race.

A large buyback could support the stock and signal confidence, but it would reduce flexibility if the memory cycle weakens. Retaining too much cash could reinforce the governance concerns associated with the Korea discount. The most credible policy would link distributions to sustainable free cash flow after strategic investment rather than the unusually high profits of one quarter.

Samsung Electronics: Memory Strength, Diversification and Internal Tradeoffs

Samsung Electronics is a broader company than SK Hynix. It produces memory and logic chips, operates a foundry, sells smartphones and consumer electronics, owns a display business and controls Harman. That diversification historically reduced dependence on any single segment. In the second quarter of 2026, however, memory profitability became so dominant that the group’s results were more concentrated than its corporate structure suggested.

Samsung reported consolidated revenue of KRW171.5 trillion and operating profit of KRW89.5 trillion for the quarter ended June 30, 2026. Revenue increased 28% from the previous quarter. The Device Solutions division generated KRW127.5 trillion of revenue and KRW89.2 trillion of operating profit, meaning the semiconductor business accounted for almost all group operating profit.

The memory business benefited from severe supply constraints, rising prices and a greater mix of server products. Samsung increased HBM4 sales and shipped HBM4E samples to major customers. Management expected robust second-half demand for server DRAM, enterprise SSDs and HBM, while acknowledging weaker demand in mobile devices and personal computers.

Long-term supply agreements signal a strategic shift

Samsung said it had signed agreements with the five largest global data-center companies and was nearing agreements with five other major customers. Management aimed to cover roughly two-thirds of long-term memory output through contracts lasting at least five years. The agreements typically included upfront payments and floor prices.

For customers, long-term agreements secure supply in a market where memory availability can constrain the deployment of expensive AI accelerators. For Samsung, deposits and price floors reduce the risk that capacity investment becomes stranded. The contracts also suggest that major cloud and AI companies view memory as a strategic bottleneck rather than a commodity purchased only in spot markets.

The arrangements do not eliminate cyclicality. Customers may negotiate flexibility, product-mix changes or volume bands. Technology can shift from one memory generation to another. AI architectures may become more efficient. Governments may impose trade restrictions. Contracts improve visibility, but their value depends on enforceability, counterparty credit and the precise pricing formula.

Record chip profits hurt other Samsung businesses

The same memory prices that enriched Samsung’s semiconductor division increased costs for its smartphone business. The Mobile eXperience and Networks businesses reported KRW33.2 trillion of revenue and an operating loss of KRW0.7 trillion. Higher component costs pressured margins even as Galaxy device sales remained solid.

This internal tension matters. Investors often view Samsung’s diversification as a stabilizer, but when one division supplies another at market-linked prices, a memory boom can transfer profit within the group rather than create equal benefits everywhere. If smartphone demand weakens because retail prices rise, Samsung can lose volume or absorb higher costs. The memory division still wins, but the group’s consumer businesses become less profitable.

The Visual Display and Digital Appliances businesses also recorded a slight operating loss, while Samsung Display generated KRW0.7 trillion of operating profit on KRW7.5 trillion of revenue. The quarter therefore revealed a highly uneven group: a world-leading memory franchise producing extreme margins alongside consumer segments struggling with the consequences of component inflation.

HBM competition is improving, but execution remains critical

Samsung entered the current HBM cycle behind SK Hynix in customer qualification and market share. By mid-2026, it had gained ground, added major customers and expected HBM4 revenue to more than triple in the third quarter. Management said HBM share could move closer to Samsung’s share in conventional DRAM during the second half.

That outcome would materially change competitive economics. HBM has higher value per bit, stronger customer collaboration and more demanding packaging requirements than ordinary DRAM. A larger Samsung presence could increase industry supply and pressure pricing, but it could also validate the depth of demand if customers absorb the output.

Investors should watch yield, qualification schedules and product mix rather than only shipment announcements. Sampling a product is not the same as mass production, and mass production is not the same as profitable volume. The timing of HBM4E and next-generation products will influence whether Samsung can close the gap without sacrificing margin.

Foundry improvement remains a separate option

Samsung’s foundry business competes with Taiwan Semiconductor Manufacturing Co. and Intel for advanced logic-chip manufacturing. The unit improved during the second quarter, supported by HBM base-die demand and U.S. customer orders. Samsung planned to ramp second-generation 2-nanometer mobile products and expand 4-nanometer low-power and base-die products.

A durable foundry turnaround would give Samsung another AI growth engine, but it should not be assumed from one quarter. Foundry success depends on manufacturing yields, design ecosystem, customer trust, advanced packaging and the willingness of chip designers to diversify away from TSMC. Samsung’s Taylor, Texas, facility and potential second fab increase strategic relevance in the United States, yet they also require substantial capital before generating returns.

Cash raises the same allocation question

Samsung ended June with net cash of approximately KRW167 trillion. Chief Financial Officer Park Soon-cheol said the company was discussing special dividends and updates to its shareholder-return program. Samsung has more financial flexibility than almost any semiconductor company, but it also has more capital demands: memory, foundry, displays, mobile devices, U.S. manufacturing and future technology platforms.

The market will judge any payout against those opportunities. A special dividend may reduce the Korea discount and reward shareholders after a record quarter. Excessive distribution could weaken Samsung’s ability to invest through the cycle. The relevant question is not whether the company can pay more; it clearly can. It is whether management can establish a predictable framework that returns surplus capital while preserving strategic capacity.

Is AI Memory a Structural Growth Story or Another Semiconductor Cycle?

The bullish case begins with architecture. Modern AI accelerators can perform enormous numbers of calculations, but their usefulness depends on moving data quickly enough to keep the processors occupied. HBM places stacked memory close to the processor and connects it through wide interfaces, increasing bandwidth and reducing energy per transferred bit. As models grow and inference workloads expand, memory capacity and bandwidth become critical system constraints.

Agentic AI may intensify those requirements. Systems that plan, call tools, maintain context and execute multi-step tasks can generate more memory traffic than a simple question-and-answer model. Training large models requires storing parameters, gradients and intermediate data. Inference requires loading model weights and managing key-value caches. Improvements in software efficiency may reduce memory needed for a given task, but lower cost can also increase usage.

The structural argument is therefore credible: AI expands the addressable market for high-performance memory, raises the value of bandwidth and deepens customer collaboration. Long-term contracts support the view that customers expect persistent demand rather than a brief inventory cycle.

The cyclical counterargument is equally credible. Memory remains a manufacturing industry with long lead times, high fixed costs and a history of synchronized investment. When prices rise, suppliers expand capacity. New entrants receive financing. Customers over-order to secure supply. Inventory builds. Once capacity arrives or demand growth slows, prices can fall rapidly because the marginal cost of producing additional bits is far below the capital cost of the factory.

HBM is less commoditized than conventional DRAM, but it is not immune. Products require customer qualification and advanced packaging, which slow supply growth. Yet Samsung, SK Hynix and Micron are all investing, while Chinese companies are pursuing domestic memory capabilities. Packaging capacity can expand. Yields can improve. Alternative architectures may reduce HBM intensity. The shortage can last longer than in previous cycles without becoming permanent.

The most reasonable interpretation is that AI has lifted the long-term demand curve while leaving the industry cyclical. Structural growth does not mean a straight line. Cloud computing structurally increased data-center demand, yet server and memory cycles continued. Smartphones structurally expanded semiconductor consumption, yet handset and component inventories still rose and fell. AI can create a larger, more profitable industry and still produce severe corrections.

The capex-to-cash-flow relationship will reveal the quality of growth

Investors should compare capital expenditure across the entire AI chain with the cash flow generated by AI services. Memory suppliers can only maintain pricing if customers continue to invest. Hyperscalers can only sustain investment if revenue and strategic value justify the spending. A quarter in which cloud companies generate less free cash flow because capital expenditure rises is not necessarily negative; infrastructure is built before revenue arrives. The concern becomes material when spending grows faster than plausible monetization for several years.

Long-term memory contracts partly transfer that risk from suppliers to customers. Upfront payments and floor prices protect manufacturers, but they may leave buyers committed to capacity that becomes less valuable if AI demand disappoints. The contracts therefore improve the earnings visibility of Samsung and SK Hynix while concentrating more economic risk in the hyperscalers.

Conventional memory still matters

The AI story is often reduced to HBM, but conventional DRAM and NAND remain major profit drivers. AI servers require standard memory and storage in addition to HBM. Data pipelines, checkpoints and retrieval systems consume enterprise SSD capacity. The buildout of data centers raises demand for networking, power management and storage throughout the system.

Samsung’s second-quarter results showed the benefit of broader exposure. Conventional memory prices rose rapidly, and Samsung’s ability to increase pricing contributed to its record margin. SK Hynix’s greater HBM exposure was strategically valuable but meant it participated less in some of the sharp conventional DRAM price increases. This explains why market leadership within memory can shift even while both companies benefit from AI.

Valuation: Why Single-Digit Multiples Can Mislead

One of the strongest bullish arguments for Korean equities is valuation. Forecast earnings for Samsung Electronics and SK Hynix rose so quickly in 2026 that forward price-to-earnings ratios fell even as share prices surged. The MSCI Korea Index showed a forward P/E of 5.27 as of July 31, 2026, compared with a trailing P/E of 17.16. That gap reveals how much profit growth analysts expected.

A low forward multiple can indicate genuine undervaluation. If earnings are durable, a company trading at five or six times next year’s profit can generate a high earnings yield and substantial cash. The market may be applying an excessive discount because of past cyclicality, governance concerns or temporary volatility.

The same multiple can also be a warning. Cyclical companies often look cheapest near peak earnings because the denominator is unusually large. When memory prices fall, forecast profit can be revised down faster than the share price. A stock that appears to trade at five times earnings can become a 15-times stock without moving if earnings estimates fall by two-thirds.

Investors should therefore ask what level of margin is embedded in the estimate. SK Hynix’s 76% operating margin and Samsung’s roughly 70% semiconductor margin were exceptional. If long-term contracts and structural scarcity preserve a large portion of those margins, current valuations may be conservative. If supply normalizes and pricing declines, peak earnings may be a poor basis for valuation.

Free cash flow provides another test. Semiconductor accounting profits can exceed cash generation during periods of aggressive investment because capital expenditure consumes cash before the new factories produce revenue. A valuation based on operating profit should be compared with cash after capital expenditure. The difference is especially important when companies announce tens of trillions of won in new capacity.

Balance sheets reduce downside risk but do not eliminate it. Samsung and SK Hynix held enormous net cash positions at the end of June. Cash can fund investment, dividends and buybacks. It can also accumulate without creating value if management allocates it poorly. Governance reform determines whether shareholders receive the benefit.

The Korea Discount: Governance, Capital Allocation and Reform

Korean equities have historically traded below comparable global companies because of what investors call the “Korea discount.” The term combines several issues: controlling-family influence, complex ownership structures, cross-shareholdings, low dividend payouts, treasury-share practices, related-party transactions, limited board independence and the possibility that management prioritizes group control over minority shareholders.

Cyclicality also contributed. Korea’s market has large weights in semiconductors, autos, chemicals, shipbuilding and heavy industry. These businesses are sensitive to global growth, capital expenditure, trade and commodity prices. Investors generally pay lower multiples for earnings that fluctuate sharply.

In recent years, the government and the Financial Services Commission pursued a Corporate Value-up agenda intended to improve capital efficiency, disclosure and shareholder returns. Reforms expanded English-language disclosure, increased transparency around treasury shares and strengthened directors’ duties toward shareholders. Revised rules also moved toward mandatory cancellation of treasury shares, with exceptions subject to shareholder approval and disclosure.

The reform story matters because valuation is not determined only by earnings. Two companies with identical profit can trade at different multiples if one has stronger governance, clearer capital-return policies and better protection for minority owners. If Korean companies increase dividends, cancel treasury shares, simplify ownership and disclose long-term return-on-equity targets, the market can command a higher multiple even without faster growth.

There are reasons for caution. Reform announcements do not automatically change corporate behavior. Controlling shareholders can comply formally while preserving influence through complex structures. Boards may remain deferential. Cash-rich companies may make acquisitions that serve group strategy more than minority investors. Investors should evaluate company-level action rather than assuming that a national program benefits every stock equally.

Treasury-share cancellation can improve per-share economics

Treasury shares are issued shares repurchased by a company and held rather than canceled. They generally do not receive dividends or vote while held, but they can later be reissued, used in transactions or deployed as compensation. Investors have criticized the Korean practice because treasury shares can be used in ways that affect control or dilute shareholders.

Permanent cancellation reduces the share count and makes the reduction irreversible. If profit remains unchanged, earnings per share increase. More importantly, cancellation signals that repurchases are intended to return capital rather than preserve a pool of shares for future corporate actions.

Dividend income creates second-order winners

Taylor’s discussion highlighted an underappreciated channel: companies that own stakes in Samsung Electronics, SK Hynix or related holding companies can benefit from higher dividends and asset values. Insurers, holding companies and conglomerate affiliates may receive more investment income when core holdings distribute record cash.

This derivative exposure is not automatically superior to owning the chip companies. Holding-company discounts can persist, and cross-holdings can create governance complexity. But it can provide a different balance of valuation, income and operating exposure. An insurer that receives a large dividend from a strategic equity stake may report higher earnings without assuming the same direct semiconductor-cycle risk.

Beyond Samsung and SK Hynix: Korea’s Broader Industrial Case

The narrowness of the KOSPI can obscure the diversity of Korea’s economy. The country is a major producer of automobiles, batteries, ships, industrial equipment, defense systems, displays, chemicals, pharmaceuticals and consumer technology. If investors focus only on the two largest chipmakers, they may miss businesses benefiting from the same global investment cycle or from unrelated structural trends.

Semiconductor equipment and materials

HBM and advanced DRAM require manufacturing equipment, testing, packaging, substrates and process-control systems. Korean suppliers can benefit from capital expenditure at Samsung and SK Hynix even if memory prices later normalize. Equipment companies often have more diversified customer bases and can sell into foundry, logic and display markets.

The risks differ from those of memory manufacturers. Equipment revenue can be lumpy, customer concentration can be high and orders can be delayed when chipmakers cut spending. Yet the companies may have higher returns on capital because they do not fund multibillion-dollar fabrication plants. Investors should examine order backlogs, service revenue, export exposure and dependence on a single process node.

Electrical equipment and grid infrastructure

AI data centers require power generation, transformers, switchgear, transmission equipment and cooling. Korean industrial companies such as HD Hyundai Electric and LS Corp. have exposure to global grid investment. This theme extends beyond AI. Electrification, renewable integration, industrial reshoring and aging infrastructure all increase demand for power equipment.

Grid-equipment companies can offer a less direct way to participate in the AI buildout. Their fortunes depend on utility capital expenditure, project execution, raw-material costs and manufacturing capacity rather than memory pricing. That diversification can be valuable when semiconductor valuations become crowded.

Shipbuilding and marine engineering

Korean shipbuilders are global leaders in liquefied-natural-gas carriers, container ships and specialized vessels. Order books can span years, providing visibility that differs from the short memory cycle. Energy security, fleet replacement and environmental regulation support demand for efficient ships and alternative-fuel designs.

The industry remains cyclical and execution-intensive. Fixed-price contracts can become unprofitable when steel, labor or equipment costs rise. Currency movements affect competitiveness. A full valuation should compare order value, margin quality and delivery schedule rather than treating a large backlog as guaranteed profit.

Defense and aerospace

Hanwha Aerospace, Hyundai Rotem and other Korean defense companies have expanded internationally as governments increase military spending and seek suppliers capable of rapid production. Korea’s manufacturing base, competitive pricing and experience with artillery, armored vehicles and aircraft create export opportunities.

Defense orders are politically sensitive and can be delayed by elections, financing or export approvals. Revenue recognition may occur over many years. The sector nevertheless broadens Korea’s market beyond consumer electronics and memory.

Autos and mobility

Hyundai Motor, Kia and Hyundai Mobis combine traditional vehicle manufacturing with electric vehicles, hybrids, software and components. Their global production networks give them access to U.S., European and Asian demand. Auto valuations have often reflected cyclical risk, capital intensity and concerns about the transition to electric vehicles.

A stronger governance framework and clearer shareholder returns could increase the value of these businesses even without semiconductor-like growth. The key indicators are pricing, incentives, regional mix, warranty costs, battery strategy and capital discipline.

Financials and insurers

Korean banks and insurers can benefit from improved shareholder-return policies, higher dividends and a stronger domestic economy. KB Financial, Hana Financial and Samsung Fire & Marine Insurance appeared among the largest holdings of the Matthews Korea Active ETF on August 4, 2026.

Financial companies remain exposed to household debt, property prices, credit quality and regulation. The Bank of Korea raised its policy rate to 2.75% in July, which can support net interest margins but also increase borrower stress. Insurers’ investment income and capital ratios depend on rates, market values and accounting treatment.

Matthews Korea Active ETF: What Its Portfolio Says About the Opportunity

The Matthews Korea Active ETF, ticker MKOR, provides a useful case study in how an active manager can approach a concentrated market. The fund seeks long-term capital appreciation and normally invests at least 80% of net assets in common and preferred shares of companies located in South Korea. Its benchmark is the MSCI Korea 25/50 Index.

As of August 3, 2026, the fund had approximately $123.99 million in assets and a gross expense ratio of 0.79%. As of August 4, Samsung Electronics represented 18.6% of net assets, while SK Hynix represented 3.3%. Samsung preferred shares added 2.9%. That positioning was materially less concentrated in SK Hynix than the broad MSCI Korea Index, where SK Hynix had represented more than 30% at the end of July.

The remaining portfolio illustrated the broader Korean thesis. KB Financial and Hana Financial were among the largest holdings. SK Square, a holding company with exposure to SK Hynix, represented 3.5%. Semiconductor-equipment company PSK, Samsung Electro-Mechanics, SK Inc., Samsung C&T, Korea Investment Holdings, Samsung Fire & Marine Insurance, Hyundai Mobis, HD Hyundai Electric and Hyundai Motor also appeared near the top.

This structure reflects several forms of diversification:

• Direct AI exposure through Samsung Electronics and SK Hynix.
• Supply-chain exposure through equipment and component companies.
• Holding-company exposure through SK Square, SK Inc. and Samsung C&T.
• Financial exposure through banks, investment holding companies and insurers.
• Industrial exposure through electrical equipment, autos, defense and shipbuilding.

The fund’s performance also demonstrates the cost of being underweight the most explosive winners during a concentrated rally. Through July 31, MKOR’s net asset value had gained 61.73% year to date, compared with 73.48% for the MSCI Korea 25/50 Index. Over one year, the fund gained 105.93% at net asset value, compared with 135.35% for the benchmark. Active diversification reduced participation in the benchmark’s surge.

That underperformance does not prove the strategy failed. A benchmark dominated by two stocks can be difficult to outperform without accepting similar concentration. The relevant test is whether active diversification protects capital and compounds returns across a full cycle. July offered partial evidence: MKOR’s NAV fell 13.66% during the month, slightly less than the benchmark’s 14.41% decline, although the fund’s market-price return was weaker because the ETF traded at a discount.

ETF market price can diverge from net asset value

On August 3, MKOR reported a net asset value of $55.03 and a market price of $54.75, a discount of 0.51%. The 30-day median bid-ask spread was 0.29%, and daily volume was 8,591 shares. These figures matter for U.S. investors because a fund can own liquid Korean stocks yet trade with limited liquidity in New York.

An ETF’s net asset value reflects the value of its holdings, while its market price reflects supply and demand for the ETF shares. During volatile periods, time-zone differences, hedging costs and limited trading volume can widen the gap. Investors using market orders may receive a price materially different from the displayed quote. This is a product-structure risk separate from the performance of Korea itself.

What U.S. Investors Are Actually Buying Through a Korea Fund

A Korea-focused ETF is not a neutral claim on the country’s gross domestic product. It is a portfolio determined by public-market capitalization, free float and index rules. A broad Korean index can become an indirect leveraged bet on global AI capital expenditure because Samsung Electronics and SK Hynix dominate the weights.

That distinction explains why Korean equity returns can diverge sharply from Korea’s domestic economy. A strong consumer quarter does not necessarily lift the index if memory stocks fall. A weak housing market may have limited immediate effect if semiconductor profits surge. The index is tied more closely to global technology spending, exports and the won than to household consumption alone.

Currency adds another layer. A U.S. investor measures returns in dollars. A Korean stock can rise in won while the dollar return is lower if the won weakens. The opposite is also true. Exporters may benefit operationally from a weaker won because foreign revenue translates into more local currency, even as U.S. investors lose on the currency conversion. Hedged and unhedged funds can therefore produce different outcomes.

Tax treatment, withholding and fund domicile also matter. U.S.-listed ETFs generally handle foreign withholding inside the fund. Direct ownership of Korean shares or depositary receipts can have different tax and trading implications. These issues depend on account type and individual circumstances and should not be inferred from price performance alone.

Korea’s Macro Backdrop: Strong Exports, Higher Rates and Financial-Stability Risk

Korea’s macroeconomic data supported the fundamental case in mid-2026. The Bank of Korea reported that real GDP increased 0.6% in the second quarter from the previous quarter and 3.7% from a year earlier. Exports and investment were major contributors, with the semiconductor cycle spilling into broader activity.

The Bank of Korea’s May forecast projected 2.6% growth for 2026, revised upward from 2.0% in February. The OECD also projected 2.6% growth. Both outlooks emphasized the strength of semiconductors while acknowledging risks from energy prices, trade policy and global demand.

Inflation complicated the picture. The Bank of Korea projected 2.7% consumer-price inflation in 2026, above its 2% target, and raised the base rate by 25 basis points to 2.75% on July 16. The decision was unanimous. Policymakers cited stronger growth, persistent inflation and financial-stability risks.

Higher rates can support the won and bank margins, but they increase financing costs for households and leveraged investors. Korea has long carried high household debt relative to income. Rising rates can therefore slow consumption, pressure property markets and increase credit risk. The central bank must balance an export-driven boom with domestic financial vulnerability.

The rate increase also occurred on the same day that regulators announced restrictions on single-stock leveraged products. The coincidence was not accidental in an economic sense. Both actions addressed forms of overheating: monetary policy responded to inflation and growth, while market regulation responded to leverage and volatility.

A strong semiconductor cycle can conceal uneven domestic conditions

Export profits do not distribute evenly across the economy. Memory companies and their employees may benefit from bonuses, capital expenditure and rising equity values, while households face higher borrowing costs and consumer prices. Small businesses tied to domestic consumption can struggle even as GDP accelerates.

This divergence is relevant to valuation. Banks may benefit from stronger nominal growth but suffer if household delinquencies rise. Retailers may not participate fully in the export boom. Construction companies can face higher financing costs. A country-level bullish thesis should therefore distinguish export champions from domestic-demand companies.

Foreign Selling Was Not a Simple Vote Against Korea

The transcript referred to enormous foreign selling, and the broad point is correct, though the precise period matters. Reuters reported that foreign investors sold $70.8 billion of South Korean equities in the first half of 2026 and nearly $110 billion by mid-July. The selling was unusually large, but much of it reflected benchmark mechanics, concentration limits, currency hedging and profit-taking after the KOSPI nearly doubled.

That interpretation is supported by what happened on July 31. Foreign investors bought KRW7.2 trillion in one day, more than twice the previous record, as prices rebounded. Investors who had reduced exposure because Korea became too large could return after the correction lowered weights and valuations.

Foreign flows can therefore create a misleading narrative. Continuous selling during a rally may look bearish even when it is driven by risk limits. Sudden buying after a crash may look bullish even when it is short covering or tactical rebalancing. Flow data are useful, but they must be interpreted with price, positioning and benchmark changes.

The emerging-market classification matters

Korea is economically advanced but remains classified as an emerging market by MSCI. That places it inside global emerging-market funds, where country weights are constrained by the size and risk limits of the asset class. A rapid rise in Samsung Electronics and SK Hynix can make Korea a dominant position in portfolios that also need exposure to China, India, Taiwan, Brazil and other markets.

Managers may sell Korea not because they expect lower earnings but because the country consumes too much of the portfolio’s risk budget. If Korea were reclassified or market-access rules changed, the investor base and benchmark structure could shift. Such changes are gradual and uncertain, but they are part of the long-term Korea discount debate.

The Bull Case for Korean Equities

1. AI memory demand remains supply constrained

Samsung Electronics and SK Hynix both described customer demand exceeding available supply. Long-term agreements, deposits and price floors suggest that customers are willing to commit capital to secure memory. If AI infrastructure investment continues, memory pricing can remain stronger for longer than in traditional cycles.

2. Earnings estimates may still understate later years

The bullish view holds that 2027 through 2029 earnings are not fully reflected in valuations. HBM4, HBM4E, advanced server DRAM and enterprise SSDs could create a multi-year product cycle. Better yields and richer product mix may sustain high margins even as capacity expands.

3. Balance sheets are exceptionally strong

SK Hynix and Samsung held large net cash positions after the second quarter. Strong balance sheets reduce refinancing risk and allow the companies to invest during volatility. They also create capacity for dividends, repurchases and strategic acquisitions.

4. The leverage unwind may have removed weak holders

Assets in leveraged products fell sharply, retail losses forced deleveraging and hedge funds reportedly reduced positions. If the most unstable exposure has been removed, future price action may better reflect earnings and long-term capital. Lower short interest and foreign buying in late July suggested that some investors viewed the correction as technical.

5. Governance reform can raise valuation multiples

Mandatory treasury-share cancellation, improved disclosure and stronger shareholder duties could reduce the Korea discount. Even modest multiple expansion would amplify the effect of earnings growth. Financials and holding companies may benefit particularly if capital-return policies improve.

6. The market offers more than semiconductors

Electrical equipment, shipbuilding, defense, autos and financials provide alternative earnings drivers. An active approach can invest in companies benefiting from AI infrastructure or Korean industrial competitiveness without matching benchmark concentration.

The Skeptical Case

1. Current margins may be peak-cycle margins

Operating margins above 70% attract supply. Samsung, SK Hynix and Micron are expanding, while Chinese competitors are investing. If capacity arrives before demand catches up, memory prices and earnings can fall rapidly. Forward P/E ratios based on peak profit can be misleading.

2. Hyperscaler economics remain uncertain

Cloud and technology companies are committing hundreds of billions of dollars to AI infrastructure. Some have experienced pressure on free cash flow. If monetization disappoints, customers may delay data centers, renegotiate contracts or prioritize efficiency. Memory suppliers are downstream from that capital-allocation decision.

3. Long-term contracts can transfer risk rather than eliminate it

Contracts improve supplier visibility, but they may cap pricing upside and create customer concentration. If counterparties face financial pressure, enforcement and renegotiation become important. Investors need more disclosure on volume, pricing floors, deposits and cancellation terms.

4. Korea remains highly concentrated

Regulators can reduce leveraged ETF activity, but they cannot quickly change the weight of Samsung Electronics and SK Hynix. A broad Korea fund will remain sensitive to two stocks. The market can experience national-level volatility from company-specific events.

5. Retail losses can affect the real economy

Large losses among households may reduce consumption and political support for market reform. If investors borrowed against homes or other assets, the effects can extend beyond brokerage accounts. Regulation introduced after losses may also become more restrictive than markets expect.

6. Governance reform may be incomplete

Rules can improve disclosure without changing control structures. Holding-company discounts, related-party transactions and conservative capital allocation may persist. Investors should demand evidence of cancellation, dividends and board accountability rather than paying immediately for policy promises.

Three Scenarios for the Korean Market

Scenario 1: Fundamentals reassert themselves

Leveraged-product assets continue to decline, volatility normalizes and foreign investors rebuild positions. HBM contracts convert into shipments, conventional memory remains tight and hyperscaler capital expenditure produces revenue. Samsung and SK Hynix generate strong free cash flow after investment and announce credible shareholder-return policies. In this scenario, valuations recover without returning to the most speculative conditions of June.

The key evidence would include stable or rising contract-backed orders, disciplined capex, improving free cash flow at major AI customers, lower ETF trading share and a broader market advance led by industrials and financials rather than only two stocks.

Scenario 2: Earnings remain strong but stocks remain volatile

Memory demand continues, but estimates and prices swing as investors debate the durability of AI spending. Leveraged products are smaller but still available overseas. Samsung and SK Hynix report excellent results that sometimes miss expectations. The KOSPI trades in a wide range, while active managers seek opportunities in supply-chain and governance beneficiaries.

This may be the most plausible near-term scenario because structural growth and cyclical uncertainty can coexist. Earnings do not need to collapse for valuation multiples to remain unstable.

Scenario 3: Supply and monetization disappoint

Hyperscalers slow investment, Chinese competition increases, HBM yields improve faster than demand and conventional memory supply expands. Long-term contracts protect some volume but pricing falls. Earnings estimates are revised sharply lower, revealing that single-digit forward multiples were based on peak profit. Korean household losses weaken domestic demand and regulation becomes more restrictive.

Evidence would include lower customer prepayments, contract renegotiation, rising inventory, falling utilization, reduced capex guidance and weaker free cash flow across the AI chain.

What Investors Should Watch Next

• Leveraged-product assets and trading share: A sustained decline would indicate that the feedback loop is shrinking.
• KOSPI volatility: Normalization from the extreme levels of June and July would make long-only participation easier.
• Foreign flows: Continued buying after the July 31 record would suggest that the correction restored acceptable portfolio weights.
• HBM4 and HBM4E qualification: Shipment volume, yield and customer acceptance will determine competitive share.
• Long-term contract disclosure: Investors need information on duration, pricing floors, prepayments, volume commitments and cancellation protection.
• Capital expenditure: Growth backed by customer commitments is healthier than speculative capacity expansion.
• Hyperscaler free cash flow: The ability of customers to fund AI infrastructure is the ultimate demand constraint.
• Shareholder returns: Samsung and SK Hynix need credible policies for dividends, buybacks and excess cash.
• Chinese competition: Progress in DRAM, HBM and domestic equipment could alter supply and pricing.
• Broader market breadth: Participation by banks, industrials, autos and defense companies would make the Korean rally more durable.
• Bank of Korea policy: Higher rates may support the currency and contain inflation but could pressure households and property.
• Corporate governance implementation: Actual treasury-share cancellations and capital-return changes matter more than announcements.

Competitive Positioning: Samsung, SK Hynix, Micron and China

The durability of Korea’s AI-memory advantage depends on competition. SK Hynix established the early lead in HBM through customer qualification, packaging execution and close collaboration with Nvidia. Samsung retained the largest overall memory footprint and used its scale in conventional DRAM, NAND, foundry and advanced packaging to regain ground. Micron Technology added a third major supplier with growing HBM capacity and strong exposure to U.S. customers.

Competition can benefit the market by increasing supply and reducing the risk that one supplier becomes a bottleneck. It can also reduce margins. Customers generally prefer multiple qualified sources so that a production problem at one company does not interrupt an entire AI platform. Once two or three suppliers can meet the same performance specification, customers gain bargaining power.

The competitive process occurs in stages. A supplier first develops a product that meets technical targets. It then sends samples, passes customer testing, reaches acceptable manufacturing yield, begins limited shipments and finally ramps profitable volume. Headlines often compress these stages into a single claim that a company has “entered” a market. Investors should distinguish qualification from scale.

SK Hynix’s advantage is strongest when HBM is scarce and customer relationships reward early execution. Samsung’s advantage is strongest when scale, integrated manufacturing and broad product coverage matter. Micron’s opportunity comes from adding capacity in a market where customers want diversification and from serving U.S. strategic priorities. The relative position can change with each generation because HBM3E leadership does not guarantee HBM4 or HBM4E leadership.

Chinese competition is a medium-term pricing risk

China has invested heavily in domestic semiconductor capability in response to export restrictions and strategic dependence on foreign technology. Memory companies such as ChangXin Memory Technologies have sought to expand DRAM output and move toward advanced products. Chinese suppliers still face equipment, yield, intellectual-property and customer-qualification challenges, especially at the leading edge. They do not need to match the leaders immediately to affect economics.

Additional Chinese capacity in conventional DRAM can pressure the lower end of the market and free established suppliers to focus on premium products. That may initially improve the product mix of Samsung and SK Hynix. Over time, however, a competitor that gains share in mainstream memory can fund research and move into server products and HBM. The historical semiconductor pattern is that competition begins in mature nodes and advances as scale, talent and equipment improve.

Export controls can slow this process but also encourage substitution. Restrictions on advanced manufacturing equipment limit China’s access to leading technology, while state support can offset weak near-term economics. Investors should monitor actual shipment quality, customer adoption and production yield rather than relying on announcements from either side of the geopolitical debate.

Packaging is as important as wafer production

HBM is created by stacking memory dies and connecting them through advanced packaging. Producing more DRAM wafers does not automatically produce more qualified HBM. The industry needs through-silicon vias, bonding, testing, thermal management, substrates and base dies. A bottleneck in any of those areas can preserve scarcity even while wafer capacity expands.

This creates opportunities for equipment and materials suppliers, but it also makes forecasting difficult. A rapid improvement in packaging yield can release more effective supply than a new fabrication line. Conversely, a packaging problem can delay revenue despite strong customer orders. Company guidance should therefore be evaluated alongside evidence from suppliers throughout the chain.

What the Episode Means for AI Infrastructure Investors in the United States

U.S. investors often separate chip designers, cloud companies and Asian manufacturers into different categories. Economically, they belong to one capital chain. Nvidia and other accelerator designers create demand for HBM. TSMC or Samsung manufacture logic chips and base dies. Samsung, SK Hynix and Micron supply memory. Hyperscalers purchase complete systems, build data centers and attempt to earn a return through cloud services, advertising, software, subscriptions and enterprise contracts.

A shortage at the memory layer can increase the cost of every accelerator system. That benefits memory suppliers but pressures buyers. If HBM prices rise faster than the revenue generated by AI workloads, cloud companies may delay deployments, optimize software or redesign systems. The supplier’s pricing power is therefore also a tax on customer returns.

This is why the strongest quarter for memory companies can coincide with weaker sentiment toward the broader AI trade. Investors may conclude that a larger share of industry profit is moving from platform companies to component suppliers. The total profit pool does not automatically grow as fast as the cost of infrastructure.

AI capex must eventually become AI revenue

Capital expenditure is not an expense on the income statement when a data center is built; it becomes an asset and is depreciated over time. Cash leaves immediately, while accounting expense is recognized gradually. A hyperscaler can therefore report strong operating profit while free cash flow falls because construction and equipment purchases accelerate.

That pattern is sustainable if the assets generate future revenue with attractive margins. It becomes dangerous if demand is overestimated, technology changes before the equipment is fully utilized or competitors force prices down. Memory suppliers receive cash earlier in the chain, but they remain exposed when customers revise future orders.

Investors should compare AI-related capital expenditure with cloud growth, backlog, utilization, inference volume and incremental revenue. No single metric captures the return. The useful question is whether each additional dollar of infrastructure is producing enough future gross profit to justify the cost of capital.

Efficiency is both a threat and a demand accelerator

More efficient models can use less memory or computing for the same task, which appears negative for hardware demand. Lower cost can also expand usage. When inference becomes cheaper, companies may deploy AI in more products, run more agents and process more data. The net effect depends on whether demand grows faster than efficiency improves.

This is a version of the Jevons effect: efficiency can increase total consumption by making a resource more useful and affordable. It is not guaranteed. Some workloads have limited economic value regardless of price. Investors should avoid assuming that every efficiency gain either destroys or multiplies hardware demand.

Risk Management Lessons From Korea’s Leverage Event

The Korean episode offers practical lessons even for investors who never trade a leveraged ETF. The first is that position size matters independently of conviction. An investor can be correct about AI memory over five years and still suffer a permanent loss if a leveraged position is liquidated during a 40% drawdown.

The second is that liquidity can disappear precisely when a portfolio needs it. A position that appears easy to trade during a rally may become expensive to exit when market makers widen spreads and every leveraged holder sells at once. Historical average volume is a poor guide to stress liquidity.

The third is that benchmark concentration creates hidden exposure. An investor may own a Korea ETF, an emerging-markets ETF, a global technology fund and individual semiconductor stocks without realizing that all four positions depend on the same AI-memory cycle. Diversification by fund name is not diversification by economic driver.

The fourth is that daily leverage requires daily monitoring. A product designed for one-day exposure can drift far from the expected long-term result. Holding it for weeks is not equivalent to borrowing at a fixed rate and buying the stock. The return depends on the sequence of daily moves.

The fifth is that regulatory risk rises after retail losses become politically visible. Governments may tighten rules rapidly, change margin treatment, suspend new products or impose trading limits. Those actions can protect future investors while reducing liquidity for existing holders.

Questions to ask before using any leveraged product

• What is the exact daily objective, and how often is exposure reset?
• Which derivatives or counterparties provide the leverage?
• How large is the product relative to the underlying stock’s free-float liquidity?
• What happened to the product during previous high-volatility periods?
• How wide can the premium or discount to net asset value become?
• What are the financing costs, expense ratio and derivative spreads?
• Could a margin call force liquidation before the long-term thesis plays out?
• Are there position limits, trading halts or regulatory changes that could affect exit liquidity?

Why Market Breadth Is the Best Test of a Sustainable Korea Rally

A healthy market does not require every stock to rise, but it generally benefits from multiple sources of earnings growth. Korea’s first-half rally became vulnerable because so much performance came from Samsung Electronics and SK Hynix. When those stocks reversed, the index had few companies large enough to compensate.

Market breadth can be measured in several ways: the percentage of stocks above long-term moving averages, the number of advancing versus declining shares, the performance of equal-weighted indexes, sector contribution and the share of total volume concentrated in the largest names. No single measure is definitive, but together they show whether capital is spreading beyond the leaders.

A broader Korean advance would include banks benefiting from better capital returns, electrical-equipment companies with global order books, shipbuilders converting backlogs into profit, defense exporters, auto manufacturers and semiconductor suppliers. It would also include smaller companies whose earnings improve because domestic demand strengthens.

Breadth matters for political reasons as well. A market perceived as enriching only two conglomerates and leveraged traders may lose public support. A market in which pension funds, households and companies benefit from dividends and a wider range of businesses can support long-term reform.

Equal-weight performance can expose concentration

A market-cap-weighted index gives the largest companies the largest influence. An equal-weighted version gives each constituent the same weight. If the cap-weighted index rises sharply while the equal-weighted index lags, leadership is narrow. If both advance, the rally is broader.

For Korea, comparing cap-weighted and equal-weighted performance can help distinguish an AI-memory surge from a national re-rating. The first can be profitable but fragile. The second would provide stronger evidence that governance reforms and industrial competitiveness are lifting the market as a whole.

Frequently Asked Questions

Why did the South Korean stock market fall even though Samsung and SK Hynix reported record profits?

The selloff reflected more than earnings. The market was highly concentrated in Samsung Electronics and SK Hynix, foreign investors had reduced positions, retail investors used margin and single-stock leveraged products were forced to rebalance as prices fell. SK Hynix also missed unusually high analyst expectations, while investors questioned the sustainability of AI spending and peak memory margins.

What is a single-stock leveraged ETF?

It is an exchange-traded product designed to deliver a multiple of one stock’s daily return, usually through derivatives. A two-times product seeks roughly twice the daily gain or loss before fees. It does not promise twice the stock’s return over longer periods because daily resetting creates path-dependent compounding.

When did South Korea allow domestic single-stock leveraged ETFs?

The regulatory change took effect on April 28, 2026, and domestic products began trading on May 27. The framework allowed up to 200% exposure to qualifying blue-chip stocks and added education and deposit requirements.

What restrictions did South Korea impose in July 2026?

Authorities temporarily suspended new single-stock leveraged product listings, banned marketing of existing products, increased investor education, strengthened premium-discount management and raised the minimum deposit for new or additional purchases to KRW30 million in cash. The deposit rule was accelerated to July 31. Additional portfolio limits and higher trading costs were later discussed.

Are Samsung Electronics and SK Hynix still fundamentally strong?

Both companies reported exceptional second-quarter results and described robust AI-server memory demand. Their balance sheets are strong and long-term contracts improve visibility. The uncertainty concerns the durability of pricing, the return on customer AI spending, future capacity expansion and whether current margins represent a cyclical peak.

What is HBM and why is it important for AI?

High-bandwidth memory stacks multiple memory layers and connects them through wide, high-speed interfaces near an accelerator. It allows processors to access data rapidly and with lower energy per bit than conventional arrangements. AI training and inference often become limited by memory bandwidth, making HBM a critical component.

Why is the Korean market so concentrated?

Samsung Electronics and SK Hynix grew to enormous market values because of their global memory leadership and AI-related earnings. Korea has many listed companies, but few approach the scale of these two. Market-cap-weighted indexes therefore assign them dominant weights.

What is the Korea discount?

The Korea discount describes the tendency of Korean companies to trade at lower valuation multiples than comparable global peers. Explanations include cyclicality, controlling-family influence, complex ownership, low payouts, treasury-share practices and weaker minority-shareholder protection. Governance reform seeks to reduce the discount.

What does the Matthews Korea Active ETF own?

As of August 4, 2026, MKOR’s largest holding was Samsung Electronics at 18.6%, while SK Hynix represented 3.3%. The portfolio also held banks, insurers, holding companies, semiconductor equipment suppliers, electrical-equipment makers, automakers, defense companies and shipbuilders.

Did active management protect MKOR during the selloff?

In July 2026, MKOR’s net asset value fell 13.66%, slightly less than the MSCI Korea 25/50 Index’s 14.41% decline. The fund had underperformed the benchmark during the preceding rally because it held less of the most dominant chip stocks. The long-term value of the approach depends on performance across a full market cycle.

Could Korean regulators completely eliminate the leverage risk?

No. Domestic restrictions can reduce participation and improve investor protection, but leveraged products remain available in foreign markets and investors can use options, futures, margin and structured notes. Regulators can reduce the size of the feedback loop, not eliminate all leveraged exposure.

What is the most important risk to the bullish AI-memory thesis?

The most important economic risk is that AI infrastructure spending fails to generate enough revenue and cash flow for customers to sustain investment. That could slow orders just as memory capacity expands, causing pricing and earnings estimates to fall.

Final Assessment

South Korea’s 2026 market episode should not be reduced to a choice between a fundamental boom and a speculative bubble. It contained both. Samsung Electronics and SK Hynix produced results that confirmed the economic importance of AI memory. Their customers sought multi-year supply, their cash balances grew and their product roadmaps moved into HBM4 and HBM4E. Those are tangible business developments, not market mythology.

At the same time, the market built a layer of leverage around those companies that was too large for the available liquidity. Domestic, Hong Kong and U.S. products multiplied daily moves. Retail borrowing increased. Foreign funds reduced exposure as index weights exceeded risk limits. When momentum reversed, the same structures that accelerated the rally accelerated the decline.

The regulatory response is likely to reduce the most immediate risk. Higher cash requirements, marketing restrictions, education, portfolio limits and tighter market-making standards can slow speculative demand and reduce premium-discount distortions. Early volume data suggest that the measures are already changing behavior. They do not solve concentration, overseas migration or the daily-reset mechanism.

The strongest bullish evidence is the combination of record operating earnings, customer commitments, constrained supply and strong balance sheets. The strongest skeptical evidence is the extremity of margins, the scale of planned capital expenditure and the dependence of the entire chain on sustained hyperscaler spending. The market’s low forward valuation is compelling only if a substantial portion of current profit proves durable.

For the Korean equity market to become more than a volatile proxy for two memory stocks, breadth must improve. Financials, industrials, defense companies, shipbuilders, automakers and equipment suppliers need to contribute. Governance reforms must translate into canceled treasury shares, clearer dividends, disciplined investment and stronger treatment of minority shareholders. Active managers may find opportunity in those second-order beneficiaries, but diversification will sometimes lag a benchmark dominated by the hottest trade.

The lesson for global investors is broader than Korea. A correct long-term thesis can still produce severe losses when expressed through excessive leverage, crowded positioning and instruments designed for daily rather than long-term exposure. Fundamentals determine value over time. Market structure determines whether investors can remain in the trade long enough to realize it.

This article is provided for general informational purposes and does not constitute financial, investment, tax, or legal advice.

Sources

• Financial Services Commission: Capital Market Rules Change Allowing the Introduction of Single-stock ETFs
• Financial Services Commission: Measures on Single-Stock Leveraged Products, July 16, 2026
• Financial Services Commission: Early Implementation of Strengthened Deposit Requirements
• Reuters: South Korea Vows to Soothe Stock Market Volatility as ETF Boom Slows
• Reuters: Big Investors Think It Might Be Time to Buy in South Korea
• Reuters Explainer: How Leveraged ETFs Drove South Korean Market Volatility
• Reuters: South Korea Stock Rout Breaks Records
• Reuters: Inside South Korea’s World-Beating Bear Market
• Reuters: Foreign Investors Dump Asian Stocks as AI Winners Become Crowded
• SK Hynix: Second-Quarter 2026 Financial Results
• Reuters: SK Hynix Record Profit Misses Elevated Forecasts
• Samsung Electronics: Second-Quarter 2026 Results
• Reuters: Samsung Sees Chip Shortage Extending to 2028
• Matthews Asia: Matthews Korea Active ETF Fund Facts, Performance and Holdings
• MSCI Korea Index Profile and Valuation Data
• MSCI Korea 25/50 Index Profile
• Bank of Korea: Real Gross Domestic Product, Second Quarter 2026
• Bank of Korea: July 2026 Monetary Policy Decision
• Bank of Korea: May 2026 Economic Outlook
• OECD Economic Outlook 2026: Korea

Affiliate disclosure: Businessfinance.news may earn compensation from qualifying actions completed through selected links on this website, at no additional cost to the reader. Affiliate relationships do not influence our editorial reporting, analysis, or conclusions.