SK Hynix: What Justifies Its Market Cap Surpassing One Trillion?
Exclusive HBM Supply, Supply Shortages, and In-Depth Analysis of Industry Changes

On April 23, 2026, SK Hynix delivered a performance report that drew worldwide attention in the semiconductor industry. As of March 31, 2026, in the first quarter of fiscal year 2026, the company achieved operating revenue of 52.58 trillion won, a year-over-year increase of 198%; operating profit reached 37.61 trillion won, up 405% YoY; net profit was 40.35 trillion won, an approximately 398% YoY increase. The quarterly sales broke the 50 trillion won mark for the first time, with an operating profit margin rising to 72%, setting a new record since the company's founding.

This financial report is not an isolated event. Just one month later, on May 28, 2026, SK Hynix's stock price continued its strong upward trend, reaching a intraday high of 2,289,000 won and closing up 2.07% (+46,375 won) from the previous day, once again hitting a new all-time high. The company's market capitalization also further increased, firmly establishing itself in the global "trillion-dollar market cap club." During the same period, Micron Technology's market cap surpassed one trillion dollars, and Samsung Electronics had previously reached the milestone—bringing the three major memory giants into the trillion-dollar valuation threshold, indicating that the valuation focus of AI infrastructure is shifting from GPUs to memory domains.

Behind this is a structural transformation driven by artificial intelligence. High Bandwidth Memory (HBM), as a key component of AI accelerators, has become the core bottleneck constraining computing power release. SK Hynix, with over a decade of strategic investment in this field, has secured a critical position in this transformation.

Performance Breakdown: The Underlying Drivers of a Fourfold Profit Surge

Structural Changes Behind the Numbers

The financial performance of SK Hynix in Q1 2026 reveals profound changes occurring across the entire memory industry.

From the revenue structure, AI-related products have become the core growth engine. The HBM product line, as the business segment with the highest gross margin, contributes significantly to the company's profits. The company disclosed that customer demand for HBM supply over the next three years has far exceeded current capacity, indicating order visibility extending into 2027 and beyond.

In terms of profit scale, an operating profit margin of 72% indicates SK Hynix has broken free from the cyclical fluctuations typical of traditional memory manufacturers. Historically, the operating margin for DRAM has ranged around 50%-60% at peak times and could turn negative during downturns. The 72% margin shatters this historical range, demonstrating that the pricing power and structural premium of high-value-added products like HBM are reshaping the industry's profit model.

Price Drivers: All Lines of DRAM and NAND Rising

The surge in profits is supported by a comprehensive increase in memory prices. According to TrendForce data, in Q2 2026, general DRAM contract prices increased by 58% to 63% quarter-over-quarter, while NAND Flash contract prices rose by 70% to 75%. Earlier, in Q1, general DRAM contract prices were revised upward to a 93%-98% increase. During the same period, NAND prices for mobile applications increased by about 100% QoQ.

Structurally, the price trends of these two categories show clear divergence. The price rally for DRAM was front-loaded, with an overall increase of 80%-90% in Q1, then slowing to over 50% in Q2, exhibiting a pattern of "oscillating upward with decelerating growth each quarter"; NAND Flash, however, demonstrated stronger resilience, with a Q2 increase surpassing 70%, and market consensus suggests prices will continue to rise steadily in the second half, with stability far exceeding that of DRAM.

The fundamental driver behind the price increases is not short-term supply-demand imbalance but the prioritization by the three major manufacturers to allocate capacity to high-margin products like HBM and server DRAM, leading to a persistent tightening of supply for general-purpose products.

Performance Comparison: From Industry Low to Historic Peak

Reviewing SK Hynix’s profit trajectory in recent years, the magnitude of this change is even more astonishing. The company posted a net loss of about 911.2 billion won in 2023, then swung to a net profit of approximately 1,979 billion won in 2024, and further grew to about 4,291.9 billion won in 2025. Over three years, the company transformed from deep losses to single-quarter net profits exceeding 40 trillion won, a growth rate rarely seen in the history of the semiconductor industry.

SK Hynix Recent Net Profit Changes

| Year | Net Profit (Won) | Trend Characteristics | | --- | --- | --- | | 2023 | -911.2 billion | Industry trough, deep losses | | 2024 | ~525.8k | Turnaround, AI demand ignited | | 2025 | ~376.1k | Rapid growth, HBM volume expansion | | Q1 2026 | ~40.35 trillion | Single quarter surpasses full-year peak |

Data source: Company financial disclosures and MarketScreener

This "V-shaped" reversal vividly illustrates how the explosion of AI is reshaping the memory industry. It should be noted that 2025 and earlier figures are annual data, while 2026 is only for a single quarter; thus, direct comparison should be cautious.

Building HBM Dominance: Market Share and Competitive Landscape

Absolute Market Leadership

In the current global HBM market, SK Hynix holds an absolute advantage. According to Counterpoint Research data, in Q4 2025, by revenue, SK Hynix accounted for 57% of the global HBM market, Samsung for 22%, and Micron for 21%. Previously, SK Hynix led in Q2 2025 with an even higher share, with the other two gradually catching up afterward.

Looking ahead to 2026, the HBM market continues to grow rapidly, with SK Hynix maintaining its leadership. Samsung, leveraging contributions from HBM3E and HBM4, has significantly recovered its market share, and Micron is actively expanding capacity. Industry forecasts suggest the global HBM market size will grow from about $35 billion in 2025 to $100 billion by 2028. Maintaining a leading share amid this rapid expansion means SK Hynix is enjoying both "market growth" and "share advantage" benefits.

Global HBM Market Share (by revenue, Counterpoint Research)

| Time Point | SK Hynix | Samsung Electronics | Micron Technology | | --- | --- | --- | --- | | 2025 Q2 | 64% | 15% | 21% | | 2025 Q4 | 57% | 22% | 21% |

Data source: Counterpoint Research

Differentiated Competition with Samsung

The competition between SK Hynix and Samsung in HBM reflects two different technological routes and market strategies. Samsung, as the largest DRAM manufacturer globally, still leads the overall DRAM market. In Q1 2026, Samsung’s share of the global DRAM market was 38%, SK Hynix 29%, and Micron 22%.

However, in the high-margin HBM segment, SK Hynix maintains a clear lead thanks to its first-mover advantage. Samsung faced challenges during certification processes; reports indicate that its fifth-generation 12-layer HBM3E product took a long time to gain NVIDIA qualification, during which SK Hynix secured major orders. Nonetheless, once Samsung enters the supply chain, its large capacity scale could still significantly impact market share in the medium to long term.

Micron’s Aggressive Pursuit

Micron’s efforts to catch up in HBM are formidable. Its HBM accounts for 26% of total memory capacity, higher than Samsung’s 23% and SK Hynix’s 18%. Micron disclosed that the ramp-up speed of HBM4 capacity is twice that of the previous HBM3 12-layer products, with significant improvements in yield. The next-generation HBM4E is expected to begin mass production in 2027, targeting NVIDIA’s Vera Rubin platform.

Additionally, Micron’s global expansion covers the US, Japan, Singapore, India, and Malaysia, with new capacity timelines extending to 2030. In the capacity race, Micron is narrowing the gap with the leaders at a more aggressive pace.

Deep Ties with NVIDIA: Strategic Logic

From Supplier to Strategic Partner

The relationship between SK Hynix and NVIDIA has long surpassed the traditional supplier-client framework. As early as 2020, SK Hynix dispatched core engineers to NVIDIA’s US headquarters, working alongside GPU engineers on chip architecture design and thermal optimization. This deep embedded collaboration has given SK Hynix a competitive edge difficult for rivals to replicate, making it the exclusive or preferred supplier of HBM3E for platforms like H100, H200, and B100.

This strategic partnership is most evident in the Blackwell platform. The Blackwell GPU continues to sell well and predominantly adopts HBM3E solutions, with NVIDIA significantly increasing orders from SK Hynix. For the next-generation Vera Rubin platform, reports indicate SK Hynix has secured over two-thirds of HBM4 orders.

Strategic Adjustment: Prioritizing HBM3E

In April 2026, SK Hynix made a notable strategic adjustment: it decided to reduce HBM4 shipment plans for NVIDIA by about 20% to 30% from the original target, while accelerating capacity deployment and supply of HBM3E.

The core reason is that NVIDIA’s Vera Rubin chips still face certain challenges in advanced process technology, yield control, and packaging, leading to delays in product introduction and a short-term demand slowdown for HBM4. SK Hynix is prioritizing HBM3E production to ensure short-term delivery to key clients and avoid over-investment in HBM4 capacity amid uncertain demand.

From a business logic perspective, this is a typical "short-term efficiency for long-term certainty" strategic choice. It also reveals a key fact: SK Hynix’s product schedule is heavily dependent on NVIDIA’s product cycle, indicating a deep coupling between the two.

Customer Concentration: An Unavoidable Structural Risk

This deep integration brings significant commercial benefits but also structural risks. Multiple Korean media reports indicate that NVIDIA accounts for the vast majority of SK Hynix’s HBM orders; in Q1 2026, NVIDIA’s revenue contribution was about 14.8%. All HBM capacity has been sold out through 2027, but high customer concentration limits SK Hynix’s bargaining power in pricing negotiations. If NVIDIA adjusts its product schedule or adopts supply chain diversification strategies, or if competitors achieve technological breakthroughs, SK Hynix risks rapid order loss.

Supply Shortage in 2027: The Underlying Logic of a Super Cycle

Quantified Gap: Data-Backed Supply-Demand Imbalance

“The supply shortage will persist until 2027”—this consensus is supported by solid data.

According to Counterpoint data, from 2026 to 2027, DRAM annual capacity must grow at about 12% to alleviate shortages, but actual growth is only around 7.5%, creating a significant gap. This indicates that supply-demand imbalance will intensify over the coming years. Major suppliers are expanding capacity, but by the end of 2027, global DRAM supply will only meet about 60% of market demand.

Goldman Sachs recently revised upward its expectations for DRAM shortages, stating that in 2026 and 2027, global DRAM will face shortages of approximately 4.9% and 2.5%, respectively, with 2026 described as “the tightest year in the past 15 years.” Samsung also warned in its recent earnings call that supply constraints are expected to continue until 2027, with demand fulfillment rates dropping to historic lows.

DRAM Supply-Demand Gap Forecast (by institution)

| Indicator | 2026 | 2027 | Data Source | | --- | --- | --- | --- | | DRAM Supply-Demand Gap | 4.9% | 2.5% | Goldman Sachs | | Capacity Growth Rate | ~7.5% | — | Counterpoint | | Demand Growth Rate | ~12% | — | Counterpoint | | Capacity Meets Demand | ~60% | ~60% | Industry estimate |

The Dilemma of Capacity Expansion Timing

The root cause of supply shortages lies in the physical time required for capacity expansion. Despite large capital expenditures by the three major manufacturers, the new wafer capacity is highly concentrated and mainly allocated to HBM, with almost no incremental capacity for NAND.

SK Hynix’s Cheongju M15X fab has accelerated mass production in 2026, providing some additional supply; the first wafer fab in the Yongin semiconductor cluster is scheduled to come online in 2027, about three months earlier than planned. One wafer fab in Yongin has a scale equivalent to six M15X fabs, making it strategically very significant.

However, before the new Yongin fab fully releases capacity, HBM supply pressure will be difficult to significantly ease. Micron’s new lines are expected to ramp up between 2027 and 2028, and Samsung’s fifth HBM fab focused on HBM is not expected to contribute until after 2028. This means that before late 2027, effective global HBM supply growth will be very limited.

Long-term Contracting: Industry Business Model Reconfiguration

The supply shortage is fundamentally changing the memory industry’s business model. Samsung and SK Hynix have decided to abandon the long-standing annual or quarterly short-term supply contracts and instead require major global tech clients to sign 3- to 5-year long-term supply agreements.

According to UBS analysis, recent memory contracts typically involve about five-year agreements, including formats like “three-year fixed pricing with two-year extension options” or “two-year fixed with three-year extension options.” Some capacities are even locked in at pre-agreed prices. It is estimated that by 2027, about 20%-30% of industry DRAM shipments will be covered by long-term agreements, with Micron around 20%, SK Hynix about 18%, and Samsung approximately 30%. More importantly, about 60%-70% of server DDR5 volumes have been locked in through long-term agreements by large cloud providers.

In negotiations for long-term supply agreements, SK Hynix has even requested higher prepayments and price floor clauses from clients. This level of bargaining power is unprecedented in the history of the memory industry and marks a fundamental shift from reactive to proactive pricing control by suppliers.

Reassessing Investment Logic: Cyclical Stock or Growth Stock?

Core Drivers of Valuation Repricing

The significance of SK Hynix’s market cap surpassing one trillion dollars lies in the market’s reassessment of the company’s role in the AI era. This repricing is driven mainly by three factors.

First, demand certainty for HBM. Continued investment in AI infrastructure ensures HBM’s sustained benefit as a key GPU component.

Second, the persistence of supply constraints. HBM’s manufacturing complexity—requiring TSV technology for vertical stacking and advanced packaging—creates multiple bottlenecks in equipment, yield, and packaging capacity. New entrants are unlikely to threaten the market in the short term. The three dominant players—Samsung, SK Hynix, and Micron—control about 89% of the market in Q1 2026 (Samsung 38%, SK Hynix 29%, Micron 22%), with little incentive for competitive expansion that would suppress prices.

Third, a fundamental shift in business model. Long-term agreements lock in capacity and price floors, partially freeing the industry from the traditional “price hikes—capacity expansion—oversupply—price drops” cycle. Industry forecasts suggest that by the end of 2026, HBM dedicated production will account for 25% of total front-end capacity, rising to 31% in 2027. This capacity shift toward high-margin HBM further reduces available capacity for conventional DRAM, making supply tightness likely to become the norm.

Unavoidable Risks and Variables

However, several risk factors warrant ongoing market attention.

First, excessive customer concentration. NVIDIA, as SK Hynix’s largest HBM customer, means any change in its procurement strategy could significantly impact SK Hynix.

Second, the pace of competitors’ catch-up. Samsung and Micron are accelerating HBM capacity and technology R&D, potentially altering the competitive landscape by 2027-2028.

Third, geopolitical and supply chain security considerations. US and EU efforts to localize semiconductor supply chains may lead to more HBM orders flowing to non-Korean manufacturers like Micron. Micron’s global expansion covers the US, Japan, Singapore, India, and Malaysia, which could influence future customer choices.

Fourth, structural pressures on downstream storage chip demand. As memory costs soar, the proportion of memory in low-end smartphones is expected to double from around 20% to 40%, possibly suppressing end-user demand.

Fifth, costs and risks associated with technological iteration and capacity adjustments. SK Hynix has already cut its 2026 HBM4 shipment plan by 20%-30%, reflecting the unpredictable pace of technological evolution. Uncertainties in yield control and customer onboarding for new products could impact short-term performance.

Conclusion

SK Hynix’s dominance in HBM is not accidental but the result of over a decade of technological accumulation and strategic focus. The company is currently riding a super cycle driven by the AI boom and supply shortages: Q1 net profit increased approximately 398% YoY, HBM market share remains leading, and its deep integration with NVIDIA has built a solid moat. The widespread adoption of long-term agreements is rewriting industry rules.

Yet, all industry advantages have a time dimension. 2027 will be a critical year—how new capacity is released, how quickly competitors catch up, the sustainability of AI capital expenditure, and the smoothness of technological iteration will jointly determine whether SK Hynix transforms from a “cyclical stock” into a “growth stock,” or reverts to the cycle.

For market participants focused on AI industry chain investment logic, SK Hynix offers a classic case of “how technological positioning translates into long-term value.” The answer lies not in quarterly financials but in every node of industry evolution over the next two to three years.