Memory Chips, Chipmakers, and the Stock Market: DRAM, NAND, and HBM Explained

Educational content only. This article is not investment advice and is not a recommendation to buy, sell, or hold Micron, Samsung, SK hynix, any other company named, or any fund or index discussed below.

"AI memory" headlines have become a fixture of financial news — stories about memory-chip prices spiking, chipmakers reporting record results, and a handful of companies racing to supply the memory that sits inside AI data centers. It's easy to read those headlines without actually understanding what a memory chip is, how it differs from the processors that get most of the attention, or why this specific corner of the semiconductor industry has a decades-long reputation for wild swings between shortage and glut. This article works through that: what DRAM, NAND, and HBM actually are and how they differ; who the major memory chipmakers are and how a memory maker differs from a company like a foundry or a chip designer; why the memory business is famously cyclical; and how that cyclicality and the current AI-driven demand story show up in the stock market. It does not evaluate whether any company, fund, or index discussed is a good investment — that isn't a question this article, or Zydhra as a platform, is positioned to answer for any individual reader.

What "memory chips" actually means

Not every chip in a computer or server does the same job. Broadly, semiconductor products are grouped into categories including logic chips (which do the actual computing — central processing units, graphics processing units, and other processors) and memory chips (which store the data and instructions those processors work with) (Congressional Research Service — Semiconductors and the Semiconductor Industry). Memory chips are themselves split into a few distinct technologies, and the differences between them matter.

DRAM (Dynamic Random-Access Memory) is the "working memory" a computer or server uses while it's actively running. It's fast to read from and write to, which is exactly why processors rely on it for data that's currently in use — the operating system, active applications, and data being processed in real time (Lam Research — What Is DRAM? (Semi 101)). DRAM is volatile: it needs a continuous power supply and constant electrical "refreshing" to hold its contents, and it loses everything the instant power is cut (Lam Research — What Is DRAM? (Semi 101)).

NAND flash is a different technology used for longer-term storage — the memory inside solid-state drives (SSDs), USB drives, and the storage in a smartphone. NAND is non-volatile: it retains data without power, which is why a phone's photos are still there after it's turned off, but it's generally slower to access than DRAM (Congressional Research Service — Semiconductors and the Semiconductor Industry).

HBM (High Bandwidth Memory) is the term generating most of the recent headlines, and it's the one most often misunderstood. HBM is not a third, separate category of memory alongside DRAM and NAND — it is a specific, standardized form of DRAM. HBM was co-developed by AMD and SK hynix, standardized by the semiconductor engineering standards body JEDEC, and works by stacking multiple DRAM dies vertically and connecting them through a very wide interface, which lets it move far more data per second than conventional DRAM modules while sitting physically close to a processor (JEDEC — High Bandwidth Memory (HBM3) DRAM standard; Wevolver — High Bandwidth Memory: Concepts, Architecture, and Applications). In practice, HBM is packaged directly alongside AI accelerator chips and graphics processors — it's the fast memory sitting immediately next to the processor doing the heavy computation, not a competing storage format to NAND. JEDEC has continued to update the HBM standard as the technology has advanced, from the original HBM specification through HBM2, HBM2E, HBM3, and now HBM4 (JEDEC — JEDEC Updates Groundbreaking High Bandwidth Memory (HBM) Standard; JEDEC — High Bandwidth Memory (HBM4) DRAM standard).

The short version: DRAM and HBM are both volatile working memory (HBM is simply a specialized, stacked, high-speed packaging of DRAM), while NAND is non-volatile storage. All three are manufactured, in large volumes, by a small handful of companies — which is where the industry structure becomes relevant to understanding the stocks tied to it.

Who actually makes memory chips — and how that differs from a foundry or a chip designer

The following describes corporate structure and US listing mechanics factually. It is not an evaluation of any company's business, and it is not an assessment of whether any listing — direct, ADR, or over-the-counter — is a more or less attractive investment.

Memory manufacturing has historically followed a different business model than the most talked-about part of the chip industry. Many advanced logic chips today are designed by "fabless" companies that don't own factories at all (Nvidia and AMD are commonly cited examples) and then manufactured by a separate contract manufacturer, or "foundry," such as TSMC (Congressional Research Service — Semiconductors and the Semiconductor Industry). Memory chips, by contrast, are still made almost entirely by integrated device manufacturers (IDMs) — companies that design and manufacture their own chips in their own factories, rather than splitting design and production between separate companies (Congressional Research Service — Semiconductors and the Semiconductor Industry). That's a structural reason memory makers, foundries, and fabless chip designers are different kinds of companies with different roles in the supply chain, even though headlines sometimes lump "chipmakers" together as if they're interchangeable.

The global memory business today is concentrated among a small number of companies:

Micron Technology is the only major memory maker headquartered in the United States, producing DRAM, NAND, and NOR memory and storage products across data-center, mobile, client, and automotive markets, according to the company's own SEC filings (Micron Technology — 10-K, fiscal year 2025, SEC EDGAR). Micron trades on the Nasdaq under the ticker MU.
Samsung Electronics, through its Device Solutions division and Memory Business unit, is one of the largest memory producers in the world by volume, spanning both DRAM and NAND, and describes itself as having led the global memory market since 1993 (Samsung Semiconductor — Memory Business Area). Samsung is headquartered in South Korea and its primary listing is on the Korea Exchange.
SK hynix, also headquartered in South Korea, is a dedicated memory maker producing DRAM, NAND, SSDs, and HBM, and traces its current corporate lineage back to Hyundai Electronics before becoming SK hynix in 2012 (SK hynix Newsroom — Fact Sheet). SK hynix's primary listing is also on the Korea Exchange.
Kioxia (Japan, spun out of Toshiba's former memory business) and Western Digital's SanDisk unit are the next-largest NAND-focused producers, and both have registered meaningful share gains during the recent AI-driven NAND demand pickup (TrendForce — AI Infrastructure Continues to Strengthen NAND Flash Demand; Kioxia Posts Highest QoQ Growth of 33.1% in 3Q25).

A notable, very recent development for US-based investors: for most of these companies' history, buying South Korean or Japanese memory makers directly meant trading on a foreign exchange or holding a thinly traded over-the-counter American Depositary Receipt (ADR). Samsung Electronics trades in the US only as an OTC ADR under the ticker SSNLF, which — as an over-the-counter listing rather than a listing on a major exchange — typically carries lower trading volume and wider bid-ask spreads than an exchange-listed stock (CNBC — SSNLF quote). SK hynix took a different path: it filed a registration statement with the SEC in June 2026 and completed an ADR offering, beginning trading directly on the Nasdaq under the ticker SKHY in July 2026. Its final prospectus priced the offering at 177,900,000 ADSs at US$149.00 per ADS — roughly $26.5 billion, below the larger size floated at the filing stage — with each ADS representing one-tenth of a common share (SK hynix Inc. — Form F-1, SEC EDGAR; SK hynix Inc. — Form 424B4, SEC EDGAR; CNBC — South Korea's SK Hynix plans $29 billion Nasdaq ADR listing). This is a factual description of a listing event, not a comment on the merits of the ADR structure or the stock.

Why the memory business is famously cyclical

The following draws on cited third-party industry data and reporting. It is not a definitive or complete account of the industry's history, and it is not a recommendation regarding any company or fund. Readers should verify current figures independently and consider speaking with a licensed financial professional before drawing conclusions.

Memory chips — DRAM in particular — behave more like a commodity than most other semiconductor products. A given generation of DRAM made by Micron, Samsung, or SK hynix is broadly interchangeable with the same generation made by a competitor, which means competition centers heavily on price and manufacturing cost rather than product differentiation. Combine that with a capital-intensive, long-lead-time manufacturing process — new fabrication capacity takes years and tens of billions of dollars to bring online — and you get a structurally mismatched market: supply expansion decisions are made years before that capacity actually ships, while demand can shift quickly. When prices rise, producers tend to invest heavily in more capacity at roughly the same time; that capacity then arrives together, tends to outrun demand, and prices fall — often sharply. This boom-bust pattern has repeated across the memory industry's history and is one of the most discussed dynamics in semiconductor investing commentary (SemiAnalysis — Memory Mania: How a Once-in-Four-Decades Shortage Is Fueling a Memory Boom).

The cycle has also reshaped who's left standing. The DRAM market had roughly 20 or more meaningful suppliers as recently as the 1980s and 1990s, including companies like Texas Instruments, Mostek, National Semiconductor, and numerous Japanese and Korean producers (EE Times Asia — Top Three Suppliers Held 94% of 2021 DRAM Marketshare). Repeated downturns forced weaker producers into mergers, government bailouts, or bankruptcy: the 1997 Asian financial crisis contributed to the merger that eventually became Hynix (later SK hynix); Elpida was formed in 1999 from NEC's and Hitachi's DRAM operations before later going bankrupt; and Micron acquired Texas Instruments' memory business in 1998 (EE Times Asia — Top Three Suppliers Held 94% of 2021 DRAM Marketshare). Consolidation has continued into the more recent past: Intel exited the NAND and SSD business entirely, selling it to SK hynix in a deal that closed in phases between 2021 and 2025 (Intel Corporation — Form 8-K, SEC EDGAR; SK hynix Newsroom — SK hynix to Acquire Intel NAND Memory Business). By 2021, the top three DRAM suppliers — Samsung, SK hynix, and Micron — together held roughly 94% of the global DRAM market (EE Times Asia — Top Three Suppliers Held 94% of 2021 DRAM Marketshare).

The most recent full boom-bust cycle before the current one played out in 2022–2023. Pandemic-era over-ordering left the industry with bloated inventories just as consumer electronics demand weakened, and memory prices fell sharply. Micron announced a roughly 20% production cut in November 2022, and SK hynix followed with its own cuts shortly after (The Motley Fool — Samsung Cuts Memory Production, and It's a Huge Win for Micron). SK hynix posted a fourth-quarter 2022 operating loss of roughly 1.7 trillion Korean won, compared with a 4.2 trillion won profit a year earlier, with the company's finance chief describing the price decline at the time as the sharpest since the 2008 financial crisis (Investing.com — Chipmaker SK Hynix reports record Q4 loss as demand weakens). Samsung, which had initially said it would not cut production, eventually reversed course and reduced output in 2023 as the downturn deepened (The Motley Fool — Samsung Cuts Memory Production, and It's a Huge Win for Micron).

These are historical, publicly reported events, not a prediction of what will happen next. Past performance and past industry cycles do not guarantee or predict future results. The point of this history isn't that a downturn is imminent or that one isn't — it's that boom conditions and bust conditions have both occurred repeatedly in this specific industry, driven by structural features (commoditized products, long capacity lead times, concentrated supplier base) that still exist today.

The AI/HBM demand dynamic — a current, fast-moving story

The following describes a recent, actively evolving industry dynamic, drawn from cited third-party sources. It is not a forecast, is not a definitive account of where the cycle currently stands, and is not a recommendation regarding any company, fund, or index. Data providers differ on specific figures, as noted below, and readers should verify current numbers independently.

Since roughly 2023, demand for AI accelerator chips (like the graphics processors used to train and run large AI models) has driven a parallel surge in demand for the HBM that sits alongside those processors, along with knock-on demand for conventional DRAM and NAND as memory makers reallocate manufacturing capacity toward HBM production (TrendForce — HBM4 Raises the Bar on Manufacturing Complexity, Premium Expected to Exceed 30%). Because a given wafer of manufacturing capacity can produce either HBM or conventional memory but not both at the same time, and because HBM is more complex and lower-yielding to produce, industry reporting has described memory makers shifting capacity away from standard DRAM and NAND toward HBM throughout 2025 and into 2026, tightening supply of conventional memory even as HBM output ramps (Tom's Hardware — AI data centers are swallowing the world's memory and storage supply).

This has coincided with a sharp rise in industry-wide memory revenue and pricing. According to TrendForce, industry-wide DRAM revenue rose 30.9% quarter-over-quarter in the third quarter of 2025, and a further 29.4% quarter-over-quarter in the fourth quarter of 2025, when Samsung reclaimed the top revenue-share position from SK hynix (TrendForce — Global DRAM Revenue Jumps 30.9% in 3Q25; TrendForce — Price Rally Drives 4Q25 DRAM Revenue Up 29.4%; Samsung Regains No. 1 Market Share). (These are past industry-wide revenue and pricing figures — not a forecast, and not an indicator of how any individual company's stock has performed or will perform.) Reporting on subsequent TrendForce data put Samsung's DRAM revenue share at roughly 38.6% in the following quarter, up from about 36.5% previously, with SK hynix and Micron holding smaller shares (TechTimes — Samsung Leads DRAM Market Share at 38.6%). NAND has moved similarly — TrendForce reported Samsung holding roughly 32.3% of the NAND market and SK hynix roughly 19.3% in the third quarter of 2025, with Kioxia (around 15.3%) and SanDisk (around 12.4%) also gaining share amid strong AI-server-driven demand (TrendForce — AI Infrastructure Continues to Strengthen NAND Flash Demand).

HBM market-share figures are a good example of a number that genuinely differs across sources, and it's worth naming that disagreement rather than picking one figure silently. SK hynix's own SEC registration statement, citing data from research firm IDC, states that SK hynix ranked first globally in the HBM market by revenue, with a 56.4% share in the first quarter of 2026 (SK hynix Inc. — Form 424B4, SEC EDGAR). Other quarterly trackers have reported different splits for different periods and using different methodologies (revenue share versus shipment or allocation share) — for example, one industry-data summary reported SK hynix at roughly 62% in a more recent quarter with Micron at roughly 21% and Samsung at roughly 17%, describing Micron as having overtaken Samsung on some customer allocations even as SK hynix retained the largest overall share (Astute Group — SK hynix holds 62% of HBM, Micron overtakes Samsung, 2026 battle pivots to HBM4). What's consistent across sources: SK hynix has held the largest share of HBM revenue since the technology's early AI-driven ramp, with Samsung and Micron both actively working to close the gap, particularly as the industry transitions toward the newer HBM4 standard (TrendForce — HBM3 Initially Exclusively Supplied by SK Hynix, Samsung Rallies Fast After AMD Validation).

None of this — the demand growth, the price increases, or any single company's HBM share — is a guarantee about what happens next. The memory industry's own history, described above, is a history of demand surges followed by aggressive capacity build-outs and subsequent price declines. Whether the current AI-driven demand cycle plays out differently than prior cycles is a genuinely open, actively debated question among industry analysts, not a settled fact, and this article is not taking a position on it.

How the memory and semiconductor sector shows up in the stock market

The following describes indexes and funds factually — what they track and how they're constructed — and is not an evaluation of their investment merit or a comparison meant to favor any of them. This is not a recommendation to buy, sell, or hold any index, fund, or security named. Cyclicality and past price behavior described below are risk characteristics, not predictions.

Semiconductor stocks, including memory makers, are tracked by a specific benchmark that shows up constantly in financial media: the PHLX Semiconductor Sector Index, commonly abbreviated SOX. SOX is a modified market-capitalization-weighted index of US-listed semiconductor companies, originally created by the Philadelphia Stock Exchange in 1993 and now maintained under Nasdaq (Nasdaq — PHLX Semiconductor Sector Index Overview; Nasdaq — PHLX Semiconductor Sector Index Methodology (PDF)). As of 2026, SOX included a mix of the largest US-listed companies across the semiconductor value chain — logic and processor designers, chip-equipment makers, and memory makers, including Micron — rather than being a memory-specific benchmark (index composition changes over time).

Retail investors more commonly encounter semiconductor exposure through exchange-traded funds built around this sector, such as the iShares Semiconductor ETF (SOXX), which tracks an index of US-listed semiconductor companies and, as of its most recent published fact sheet, has included Micron among its larger positions alongside processor and equipment makers (iShares — SOXX Fund Fact Sheet (PDF)), and the VanEck Semiconductor ETF (SMH), which tracks the MVIS US Listed Semiconductor 25 Index, a benchmark of the 25 largest US-listed semiconductor companies (VanEck — SMH Semiconductor ETF). Both are broad semiconductor funds spanning memory, logic, and equipment companies together — neither is a memory-specific or "AI memory" fund, and this article isn't evaluating either one. Investors curious about how a narrower, single-theme technology ETF is put together — as opposed to a broad sector fund like these — may find our explainer on the Defiance Quantum ETF (QTUM) useful as a comparison of methodology and concentration risk, even though it tracks a different theme entirely.

The way individual memory makers trade in US markets varies by company, and it changed meaningfully in mid-2026, as described above: Micron trades directly on the Nasdaq under MU and has done so for decades; Samsung trades in the US only as a lower-volume, over-the-counter ADR (SSNLF); and SK hynix completed a new ADR offering and began trading directly on the Nasdaq under the ticker SKHY in July 2026 (CNBC — South Korea's SK Hynix plans $29 billion Nasdaq ADR listing; CNBC — SSNLF quote). That distinction — whether a stock trades on a major US exchange, as an OTC ADR, or not at all in US markets — affects trading liquidity and accessibility, independent of anything about the underlying business.

Because memory revenue and profit swing with the boom-bust pricing cycle described earlier, memory-maker stocks have historically been described in financial commentary as more volatile than the broader market or even than the broader semiconductor sector — profits (and losses) can move dramatically within a single year as prices for a largely commoditized product rise or fall. That volatility is a structural, well-documented risk characteristic of this specific corner of the sector; it is not, and should not be read as, a prediction of future price movement in either direction. Past performance, including past cycles of sharp gains and sharp losses in memory-maker stocks, does not guarantee or predict future results. Investors who study a stock's historical price behavior sometimes use technical indicators as one input in that process; our explainer on RSI vs. the Stochastic Oscillator covers how two commonly used momentum tools are actually calculated, without suggesting either indicates what a stock will do next.

What makes this sector risky and volatile for investors to understand

Pulling the pieces above together, a few structural features are worth understanding specifically about memory-chip and broader semiconductor investing, independent of any individual company:

Commodity-like pricing dynamics. Because memory products from different makers are often close substitutes for one another, prices — and therefore memory-maker revenue and profit — can swing sharply with supply and demand, in both directions.
Long, capital-intensive build-out cycles. New manufacturing capacity takes years and enormous capital investment to bring online, meaning supply decisions made during a boom can arrive years later, potentially after demand conditions have already shifted.
Supplier concentration. A small number of companies account for the large majority of global memory production, meaning the sector's fortunes are tied to the capacity, technology, and capital-allocation decisions of just a handful of firms.
Contested, fast-moving data. As shown above with HBM and DRAM market-share figures, different research firms report different numbers for the same period, and figures that are accurate today can be outdated within a single quarter.
Cross-border listing structures. Two of the three major memory makers are headquartered outside the US, and how (or whether) their shares are accessible to US investors — direct listing, ADR, or OTC ADR — has itself been changing recently, which affects liquidity independent of the underlying business.
Narrative risk around demand cycles. The industry's own history shows repeated cycles where a period of surging demand and pricing was followed, sometimes years later, by an oversupply-driven downturn. A current demand story being real and well-documented, as the AI/HBM dynamic described above appears to be, is not the same as a guarantee that historical cyclical patterns won't recur in some form.

What this means if you're researching memory-chip stocks

Understanding how DRAM, NAND, and HBM actually differ, who makes them, and why this industry has a decades-long history of boom-bust cycles is context — it helps you read the next "AI memory" headline with a clearer sense of what's actually being reported and why the numbers move around so much between research firms and quarters. It is not, by itself, a signal about what to do with any position or how any stock will perform.

This article deliberately hasn't taken a view on whether the current AI-driven demand cycle will play out differently than the memory industry's past cycles, whether any specific company's market-share position is durable, or whether any fund, index, or individual stock discussed here fits a given portfolio. Those are questions that depend on an individual investor's own research, time horizon, risk tolerance, existing holdings, and circumstances — factors only that person, or a licensed financial professional working with them, can properly weigh. If you're researching this sector, consider reviewing companies' own SEC filings and investor disclosures directly, comparing figures across more than one research provider given how often they disagree, and, if appropriate, speaking with a licensed financial or tax professional about how any of this fits your own situation.

Disclaimer: This article is for general educational and informational purposes only. It is not investment advice, and nothing in it should be construed as a recommendation to buy, sell, or hold Micron, Samsung, SK hynix, Kioxia, Western Digital/SanDisk, any other company named, or any index, ETF, or security discussed above. Zydhra is a portfolio-tracking and analytics tool — it is not a broker-dealer and is not a registered investment adviser. Market-share figures, revenue figures, pricing data, index composition, ETF holdings, and listing details discussed above change frequently, were current as of the dates and sources cited, and may no longer be accurate by the time you read this — different research firms have also reported different figures for the same periods, as noted throughout. Past performance, including past cyclical patterns in the memory-chip industry, does not guarantee or predict future results. Before making any investment decision, consider doing independent research and consulting a licensed financial professional.