DRAM ETF Issuers: Three Giants' Market Caps Surpass Trillion, the AI Era for Storage Chips Has Only Just Begun

Original Authors: Dave Mazza, Thomas DiFazio

Original Compilation: TechFlow

Foreword: The combined market capitalization of the world's top three memory chip manufacturers has surpassed $1 trillion, prompting Morningstar to remind investors not to overlook fundamentals. Roundhill Investments (issuer of the DRAM ETF) counters point by point: AI infrastructure has reshaped the memory industry's supply-demand structure, the manufacturing barriers of HBM keep new entrants out, and the trio's expected combined profit for 2027 will reach $704 billion. Please note that the author manages the DRAM ETF and is inherently bullish.

The world's top three memory chip makers—Samsung Electronics (005930 KS), SK Hynix (000660 KS), and Micron (MU)—have each surpassed a $1 trillion market cap, joining an extremely rare club. However, this milestone has also invited scrutiny.

Morningstar recently published a blog post cautioning memory ETF investors not to ignore fundamentals, raising several pointed questions:

• History is a warning: The memory industry has repeatedly experienced boom-and-bust cycles, and investors may be ignoring this history.
• Memory companies lack an economic moat: Memory is fundamentally a commodity business, where new capacity can always enter the market and erode pricing power, offering companies little real protection for profit margins.
• The rally may be momentum-driven, not fundamental-driven: The enthusiasm surrounding memory stocks reflects excitement over AI rather than a sober analysis of profits, margins, and supply-demand dynamics.
• Valuations have surged: Memory stocks have seen significant price appreciation, potentially outpacing fundamentals.

Figure: Overview of the memory chip industry

Roundhill’s stance is: This time is different. To understand the future of the memory industry, one must first look back at its past.

History is indeed a warning, but is it still relevant?

The boom-and-bust cycle of memory chips is a fact. The most classic example occurred in the mid-1990s. When Microsoft released Windows 95 in August 1995, PCs transformed from business essentials to consumer goods, and the DRAM capacity per computer quadrupled from 1-2 megabits to 4-8 megabits. Manufacturers, caught off guard by the sudden demand surge, built factories frantically to expand capacity, ultimately leading to oversupply and a price collapse.

A similar story played out in the mid-2010s. When Apple released the iPhone 7 and upgraded the base storage from 16GB to 32GB, it seemed like a minor change, but scaling it led to a massive demand spike. Manufacturers invested heavily once again, followed by another cycle of oversupply and falling prices.

These cycles share a common pattern: Technological breakthrough → Demand surge → Capacity expansion → Oversupply → Price collapse.

The question is: Does this model still apply today?

The memory chip industry has undergone structural changes. Memory demand is no longer tied to the replacement cycle of consumer electronics but is instead linked to the computing power expansion of AI infrastructure. The scale of this market far exceeds a smartphone upgrade wave, and its growth potential is significantly greater.

DRAM and NAND prices have more than quintupled since January 2024. Hyperscalers have begun demanding long-term supply agreements to lock in bandwidth. Historically, long-term supply agreements in the memory industry were loose frameworks that changed with market conditions. But this model has shifted. SK Hynix stated during its January 2026 earnings call that current agreements reflect a "strong mutual commitment" between customers and suppliers, due to the high capital intensity of cutting-edge memory manufacturing. Micron has reported similar long-term agreement terms.

Figure: DRAM and NAND price trends

The Moat of Memory Chips: Manufacturing Complexity

Not all memory chips are the same. The memory powering today’s AI systems is called High Bandwidth Memory (HBM), which is entirely different from the memory used in phones and computers. HBM is specifically designed for AI workloads, and its manufacturing conditions are extremely demanding.

Data from Goldman Sachs shows that SK Hynix, Samsung, and Micron control nearly the entire global HBM supply. This industry, consolidated over decades, has accumulated manufacturing expertise that cannot be replicated overnight. Manufacturing complexity itself is the moat, and it is precisely what has enabled these three companies to reach their current positions.

Figure: Global HBM market share distribution

This logic is entirely different from the old cycle. In the past, it was demand rises → new capacity enters → prices collapse. The current bottleneck is not capital or willingness but technical capability. SK Hynix currently controls approximately 58% of the global HBM supply. On June 2nd, it announced plans to double wafer capacity over the next five years while warning that supply shortages will persist until 2030. Building a new factory takes at least three years, and five years or more for a greenfield site.

Furthermore, ASML—the world's sole manufacturer of extreme ultraviolet (EUV) lithography machines, essential for producing cutting-edge memory chips—entered 2026 with a backlog of orders worth 38.8 billion euros, exceeding its projected annual sales. The delivery cycle for a single EUV machine exceeds 12 months. This bottleneck cannot be resolved in the short term.

Fundamentals: Memory Manufacturers Poised to Join the World's Most Profitable Companies

The earnings, revenue, and margin expectations for Samsung, SK Hynix, and Micron reflect the secular wave of AI adoption. According to Bloomberg consensus estimates, these three companies are expected to be among the world's ten most profitable companies by 2027.

Figure: 2027 global most profitable company ranking forecast (Bloomberg consensus estimates)

The combined bottom-line profit of the three companies is expected to reach $704 billion by 2027, with total revenue exceeding $1 trillion.

Figure: Revenue forecasts for the three major memory manufacturers

Figure: Profit forecasts for the three major memory manufacturers

In terms of profit margins, the gross operating margins of Samsung, SK Hynix, and Micron have reached historical records, surpassing the previous highs seen in 2018.

Figure: Historical gross margin trends for the three major memory manufacturers

These figures are unprecedented in the history of the memory industry. Even if growth slows, as generative AI continues to integrate into the global economy, the memory industry is poised to stabilize at an unprecedentedly high baseline.

Value Reassessment in the New Profitability Era

The historic stock price performance, coupled with significant upward revisions in fundamentals, indicates that the industry is undergoing a major revaluation driven by earnings growth and margin expansion.

SK Hynix and Samsung are two typical examples. Over the past decade, the NTM (next twelve months) price-to-book ratios of the two stocks have fluctuated within a certain range, constrained by the profit characteristics of the memory industry's boom-and-bust cycles. However, this ceiling may no longer apply. The expected ROE (Return on Equity) for both companies has surged to levels unprecedented in memory industry history, necessitating a reassessment of the valuation framework investors have long used to judge these stocks.

Figure: SK Hynix NTM P/B ratio and ROE trends

Figure: Samsung Electronics NTM P/B ratio and ROE trends

Despite the astonishing recent stock price gains, the median NTM P/E ratio of the DRAM ETF holdings is only 8.37x, making valuations attractive compared to the broader tech sector. Meanwhile, the median current fiscal year earnings per share (EPS) growth rate for the portfolio is 632%. Calling memory stocks overvalued essentially applies old data to a new industry. In Roundhill's view, the gap between historical valuation conventions and current fundamental performance represents the opportunity.

Figure: Overview of DRAM ETF holdings valuation and earnings growth rate

Conclusion: Why Roundhill Isn't Worried

It is reasonable to be skeptical of a massive stock price surge. Fundamentals always matter over the long term. But in this case, fundamentals are precisely the reason memory stocks are rising.

The characteristic of the old cycle was: a demand explosion with no upper limit, leading to excessive capacity expansion by manufacturers and an inevitable price crash. Today's situation is structurally different. Manufacturing barriers limit new entrants, industry leaders themselves say supply shortages will persist until 2030, and the profit cycle is just beginning to reflect the scale of AI infrastructure buildout.

Roundhill believes that the market is currently pricing in not a bubble, but an industry that has struggled with boom-and-bust cycles for decades entering a new era.

⚠️ Editor's Note: The original authors of this article, Dave Mazza and Thomas DiFazio, are members of Roundhill Investments, which issues and manages the DRAM ETF (Roundhill Memory ETF). The article is inherently bullish. Readers should consider third-party views, such as those from Morningstar, for a comprehensive judgment. The ETF risk disclosures and legal disclaimers from the original text have been omitted. Please refer to the original article link for complete information.