April 23, 2026, marked a watershed moment for South Korea’s semiconductor industry. SK Hynix released a quarterly earnings report that redefined industry expectations: first-quarter revenue reached KRW 52.58 trillion, up 198.1% year-over-year; operating profit soared to KRW 37.61 trillion, a staggering 405.5% increase; net profit hit KRW 40.35 trillion, up 397.6%. All three core metrics—revenue, operating profit, and net profit—set new all-time records. This was also the first time a Korean chipmaker surpassed KRW 50 trillion in quarterly revenue.
Almost simultaneously, the other two major players in the global memory chip industry—Samsung Electronics and Micron Technology—were experiencing their own historic milestones. On May 26, 2026, Micron’s stock surged 19.29% in a single day, pushing its market capitalization past the $1 trillion mark (approximately $1.023 trillion), making it the third semiconductor company worldwide to join the trillion-dollar club. As of May 28, Micron’s share price stood at $928.41, with after-hours trading dipping slightly to around $904. Samsung, meanwhile, further extended its lead over SK Hynix in the DRAM market, capturing a 38% share in the first quarter of 2026.
Source: Google Finance
All three companies now find themselves at the epicenter of the AI memory supercycle. This is a story about HBM (High Bandwidth Memory)—a stacked DRAM architecture designed to interface directly with GPUs and AI accelerators, now the most sought-after strategic resource in AI infrastructure.
How HBM Became Central to the Industry
HBM isn’t a new technology, but its strategic importance has skyrocketed in the era of AI large models. Compared to traditional DDR or LPDDR memory, HBM vertically stacks DRAM chips and connects them via through-silicon vias, delivering several times the bandwidth per unit area versus conventional solutions. This enables GPUs to process massive datasets with lower power consumption, directly impacting AI training and inference efficiency.
To understand today’s competitive landscape, it’s crucial to trace a clear timeline of technological evolution and strategic differentiation:
2013–2022: Early Deployment and Strategic Divergence
SK Hynix launched the world’s first HBM product in 2013 and maintained its leadership through subsequent generations—HBM2, HBM2E, and HBM3. While Samsung has long held the top spot in global DRAM capacity and revenue, it entered the HBM segment later, only ramping up investments in the HBM3/HBM3E era. Micron skipped HBM3 entirely, jumping straight to HBM3E with a leapfrog strategy to gain late-mover advantage.
2023–2024: The HBM3E Era Takes Shape
With the generative AI boom, HBM demand surged. SK Hynix became Nvidia’s top-priority HBM3E supplier, delivering products first for the Blackwell platform. According to Counterpoint Research, SK Hynix maintained a market share above 50% for HBM from Q4 2024 through Q3 2025. In September 2024, SK Hynix became the first globally to mass-produce HBM4, further cementing its lead.
Second Half of 2025 to Early 2026: HBM4 Competition Intensifies
At CES in January 2026, Nvidia unveiled its next-generation Vera Rubin GPU architecture. HBM4 specifications were finalized: memory interface width doubled from 1,024 bits to 2,048 bits, single-stack capacity increased to 48GB (16-layer stack), and system-level bandwidth reached 22TB/s—about three times that of early Blackwell systems. At GTC 2026 in March, Samsung publicly showcased HBM4E samples, achieving 16Gbps per pin and a total bandwidth of 4.0TB/s.
Q2 2026: The Trillion-Dollar Era Arrives
By May 2026, all three memory giants had either surpassed or neared the $1 trillion market cap milestone. SK Hynix crossed $1 trillion in late May, becoming Korea’s second listed company to reach this landmark. Micron soon followed. Samsung continued to lead in market capitalization on Korea’s stock exchange.
Three Dimensions of Competitive Positioning
Market Share: Dual Structure of DRAM and HBM
To grasp the competitive dynamics among the three companies, it’s essential to distinguish between two levels of market share: overall DRAM and the HBM segment dedicated to AI.
Counterpoint Research reported on May 27, 2026, that global DRAM revenue for Q1 2026 reached $97 billion—a record high—up 80% quarter-over-quarter and 260% year-over-year. Market share breakdown:
| Metric | Samsung | SK Hynix | Micron |
|---|---|---|---|
| DRAM Market Share (Q1 2026) | 38% | 29% | 22% |
| HBM Market Share (2026 forecast) | ~28% | ~50% | ~22% |
Sources: DRAM share from Counterpoint Research; HBM share from TrendForce forecast (SK Hynix ~50%, Samsung ~28%, Micron remainder).
The mismatch between these two sets of data is notable: Samsung leads with 38% of the overall DRAM market, but SK Hynix dominates HBM—a high-value segment—with about 50% share. This reflects two divergent strategies: Samsung pursues scale and broad product coverage, while SK Hynix concentrates resources on the high-margin AI memory segment.
Quarterly shifts in DRAM market share also signal important trends: Samsung’s share increased by 2 percentage points (from 36% in Q4 to 38% in Q1), while SK Hynix’s dropped by 3 points (from 32% to 29%), indicating Samsung’s aggressive capacity expansion.
Financial Performance: Divergence in Profitability
Q1 2026 financials highlight the distinct profit profiles of the three companies during this supercycle:
SK Hynix: Quarterly revenue of KRW 52.58 trillion, operating profit of KRW 37.61 trillion, operating margin at 72%, and net margin at 77%. All core metrics set company records. DRAM ASP rose about 65% quarter-over-quarter, NAND ASP up roughly 75%. Gross margin and EBITDA margin both reached 79%. Q1 net non-operating profit was KRW 14 trillion, including KRW 1.6 trillion in net forex gains and KRW 9.9 trillion in investment asset revaluation—both non-recurring items.
Samsung Electronics: Samsung’s business spans memory chips, foundry, smartphones, appliances, and more; storage segment profits are not disclosed separately. Nomura raised Samsung’s target price from KRW 340,000 to KRW 590,000 on May 18, forecasting that AI inference demand will ignite a new memory cycle. While the market is highly optimistic about Samsung’s storage segment, overall valuation is weighed down by non-memory divisions.
Micron Technology: For FY2026 Q3 (ending May 2026), Micron guided revenue of about $33.5 billion, up over 260% year-over-year. FY2026 Q2 non-GAAP EPS was about $8.42. All HBM supply for 2026 has been sold out.
Comparing margins: SK Hynix’s 72% operating margin leads the industry. KB Securities forecasts SK Hynix’s full-year 2026 operating margin at 78.1%, potentially the highest in the global semiconductor industry—surpassing Nvidia and Saudi Aramco.
Valuation Framework: Evolving Target Prices and Pricing Logic
The valuation logic for these companies is shifting fundamentally—from "cyclical pricing" to "structural re-rating." As of May 28, 2026, here’s a summary of major broker target prices:
| Company | Broker | Target Price | Core Logic |
|---|---|---|---|
| SK Hynix | Nomura (May 17) | KRW 4,000,000 | Exponential growth in AI memory demand |
| SK Hynix | KB Securities (May 15) | KRW 3,000,000 | 2026 operating margin of 78.1% |
| SK Hynix | Mirae Asset Securities (May 27) | KRW 3,800,000 | ROE improvement and valuation re-rating |
| Micron Technology | UBS (May 26) | $1,625 | LTA-locked profits + re-rated valuation model |
Sources: Nomura report May 17; KB Securities May 15; Mirae Asset Securities May 27; UBS May 26.
Nomura’s pricing logic is the most aggressive: its target price of KRW 4,000,000 is a sharp increase from the previous KRW 2,340,000, citing "exponential growth" in AI memory demand. Cloud giants are signing long-term agreements (LTAs), locking in prices and capacity with prepayments.
UBS’s re-rating of Micron is methodologically significant. Analyst Akuri abandoned traditional sum-of-the-parts valuation, opting for a forward P/E framework based on discounted future earnings. The core assumption: even in downturns, Micron can maintain stable profitability, warranting a shift from low multiples typical of memory makers to higher levels. With profits locked in via LTAs, UBS assigns a 15x forward P/E, setting the target at $1,625.
Dissecting Market Sentiment: Four Main Lines of Debate
Market views on the three memory giants are far from uniformly bullish. The bull-bear divide centers on four main themes:
Theme 1: The Supercycle Has Arrived vs. Cyclicality Never Dies
Bulls argue for a structural shift in demand. Nomura introduces the "new mechanism" concept, claiming the old cyclical model can’t explain current supply-demand dynamics—AI memory demand is on an exponential growth path, while supply is constrained by wafer fab construction cycles, EUV lithography bottlenecks, and advanced packaging capacity limits, making expansion extremely slow.
Morningstar offers a cautious perspective: SK Hynix shares are up about 88% year-to-date in 2026, and even record-breaking quarterly data may not drive further price gains. The key question is whether the memory industry’s boom-bust cycle has been permanently eliminated or merely temporarily dormant. On the day SK Hynix released its earnings, the stock actually fell, reflecting that expectations had already been priced in.
Theme 2: SK Hynix Leads HBM vs. Samsung’s Strong Comeback
Mainstream opinion holds that SK Hynix has a clear first-mover advantage in HBM. According to Korea Economic Daily (March 2026), SK Hynix secured about 70% of Nvidia Vera Rubin’s initial HBM4 orders, with Samsung taking roughly 30%. TrendForce forecasts SK Hynix will continue to lead global HBM supply in 2026, with about 50% of HBM bit output.
But Samsung’s comeback shouldn’t be underestimated. In March 2026, Samsung passed Nvidia’s HBM4 certification (for 10Gbps and 11Gbps speeds) and began HBM4 shipments in February. Samsung has compressed its HBM R&D cycle from about two years to less than one, aligning closely with Nvidia’s annual product iterations. At GTC in March 2026, Samsung showcased HBM4E samples achieving 16Gbps per pin and 4.0TB/s total bandwidth.
Theme 3: Micron Marginalized vs. Undervalued Opportunity
A widely discussed report from Korea Economic Daily (March 2026) noted that only Samsung and SK Hynix were listed as HBM4 suppliers for Nvidia’s Vera Rubin flagship platform, with Micron absent. However, industry analysts point out that Micron may still supply HBM4 for lower-tier Rubin AI inference accelerators (such as Rubin CPX), so it’s not entirely sidelined.
UBS offers a contrasting view: Micron’s valuation starts far below its Korean peers, and once HBM business scales up and profits stabilize, its re-rating potential is greater. Micron’s entire 2026 HBM supply is sold out; HBM market size is projected to grow from $35 billion in 2025 to $100 billion by 2028, with an annual growth rate of around 40%.
Theme 4: Long-Term Agreements Reshape Valuation vs. Cyclicality Will Return
UBS research notes that new LTAs typically last 3–5 years, locking both volume and price, with buyers committing to prepayments and capex support. Cloud giants have locked in 60–70% of industry server DDR5 shipments via enhanced LTAs, and by 2027, 20–30% of industry DDR bit shipments will be covered by such agreements.
This trend is fundamentally changing how the market values memory companies. Traditionally, memory makers are seen as highly cyclical stocks—profits balloon at peaks and shrink or turn negative in downturns. If LTAs truly smooth profits, memory giants could maintain relatively stable earnings even in down cycles, supporting higher valuation multiples.
Industry Impact Analysis: Structural Transformation of the Memory Sector
The AI memory supercycle is not just a price boom—it’s fundamentally reshaping the global memory chip industry in four ways:
Structural tilt in capacity allocation. HBM die sizes are much larger than equivalent-capacity standard DRAM, meaning every HBM chip produced displaces several times its capacity in regular DRAM output. Counterpoint Research notes that since late 2025, cloud providers have accelerated AI infrastructure buildout, with surging HBM demand crowding out standard DRAM capacity and pushing the entire DRAM market into shortage. SEMI China President Feng Li points out that even with 70% of new/adjustable capacity redirected to HBM by the three majors, the HBM capacity gap remains at 50–60%. This displacement effect is key to understanding the broad rise in memory prices.
Deepening customer relationships. Unlike the standardized products of traditional memory, HBM supply chains are evolving into highly customized strategic partnerships. SK Hynix’s "One-Team" alliance with TSMC, Samsung’s synchronized R&D cycles with Nvidia, and Micron’s LTAs with cloud giants—all represent deep technical coupling and commercial lock-in. These relationships create high entry barriers for HBM, further solidifying incumbent suppliers’ competitive positions.
Accelerated R&D race. Samsung’s move to compress HBM R&D cycles to less than a year signals an industry-wide "arms race." AI chip makers are adopting annual product line updates; HBM suppliers unable to match this pace risk losing key customer orders. This acceleration means R&D intensity will keep rising, and companies with greater scale and deeper technical expertise will widen their competitive advantage.
China’s breakthrough efforts. Beyond the global triopoly, Chinese memory makers are rapidly catching up. ChangXin Memory has achieved a 1β nanometer DRAM process breakthrough and is mass-producing DDR5; Q1 2026 DRAM revenue grew over 700% year-over-year, with global market share rising from 3% to 8%, making it the world’s fourth-largest DRAM supplier. While Chinese firms still lag the big three in HBM, their capacity expansion and technological progress warrant ongoing attention.
Conclusion
Within the grand narrative of the AI memory supercycle, each of the three memory giants occupies a distinct competitive position: SK Hynix has built first-mover advantages and technical barriers in the most valuable HBM segment; Samsung maintains leadership in scale with the most comprehensive memory product lineup and largest capacity base; Micron, starting from a lower valuation and with the most aggressive re-rating expectations, offers investors a different risk-reward profile.
Ultimately, investment value may not hinge on who wins the supply race for a particular generation of HBM, but on who can sustain forward-looking technology, disciplined capacity expansion, and sticky customer relationships as AI infrastructure evolves over the long cycle. The interplay of these three capabilities will ultimately define each company’s value boundaries.
