Micron Technology Inc · Moat analysis

Report previewMU.USReport date Last updated
← Back to reports
Report contents
01

Industry landscape

Where the company competes: three DRAM suppliers set the price

Micron sells memory. The business that drives its results is DRAM, the working memory that sits next to a CPU or GPU and holds data a processor needs right now; NAND flash, which stores data when the power is off, is the second leg. Both are sold as standard parts defined by industry specifications, which tells you most of what you need to know about the competitive arena: customers can and do qualify multiple suppliers, and price moves with industry-wide supply and demand.

The value chain is short and lopsided. Micron buys silicon wafers, photoresist and other materials, plus capital equipment from a handful of vendors that includes ASML for EUV lithography and Applied Materials, Lam Research and Tokyo Electron for deposition and etch. It designs and fabricates chips in its own fabs in Japan, Singapore, Taiwan and the United States, then sells to hyperscalers (directly and through server OEMs), PC and smartphone makers, automotive and industrial customers, and distributors that resell into the channel. Micron reports revenue across four units, Compute and Networking, Mobile, Embedded, and Storage, which map to end markets; the split between DRAM and NAND cuts across all four.

The specific submarket that matters most is high-bandwidth memory. HBM stacks DRAM dies vertically, connects them through silicon vias and ships them in fixed configurations alongside an AI accelerator. It is the one part of the memory market where a supplier negotiates allocation with a named customer years ahead of shipment instead of taking a spot price. Commodity DDR5 for servers and LPDDR5 for phones remain the volume base, and both are priced by the balance of industry bit supply against demand.

Competition is global. Bits cross borders freely and the contract and spot prices that set Micron's revenue are struck in a worldwide market, so a fab in Hiroshima competes with one in Icheon or Xian on cost per wafer, not on address. Two caveats shape the boundary. Manufacturing location is not footloose: fabs are sited where governments pay for them, and US, Japanese and Indian subsidies have pulled capacity toward those jurisdictions. And export controls on China, plus tariffs, have begun to segment what was a single market.

Market structure: a three-firm DRAM oligopoly above a crowded NAND market. Public estimates of revenue share put Samsung in the low-to-mid 40s percent of DRAM, SK hynix near 30% and Micron in the low-to-mid 20s, with the three supplying close to the entire market. NAND is looser: Samsung leads at roughly 30%, SK hynix including Solidigm holds about a fifth, and Kioxia, SanDisk and Micron each sit in the low double digits, with China's YMTC adding capacity outside the leading group.

SupplierDRAM share (est.)NAND share (est.)Position in HBM
SamsungLow-to-mid 40s %~30%Qualifying; catching up
SK hynix~30%~20%Largest HBM supplier
MicronLow-to-mid 20s %Low double digitsThird entrant, ramping
Kioxia—Low double digitsNone
SanDisk—Low double digitsNone

Shares are estimates and shift with quarterly bit shipments and pricing; treat the ordering as reliable and the decimals as not. On the practical test of whether the leaders fit on one hand, DRAM passes cleanly and NAND does not. That difference matters for how the two halves of Micron should be valued.

Entry barriers are high in DRAM and moderate in NAND. A leading-edge DRAM fab costs several billion dollars, takes years to build and qualify, and requires process know-how that only three firms have accumulated at scale. Patents and trade secrets reinforce the cost curve, since the low-cost producer sets the price and everyone else earns the difference between their cost and that price. NAND has more credible entrants because the design and manufacturing problems are less concentrated, though the capital requirement still limits the field to a half-dozen.

The unresolved question is whether the DRAM oligopoly is a durable feature or a phase. China's state-backed CXMT has been adding commodity DRAM capacity, and any volume it brings lands on the standard parts Micron sells in its largest unit. The 10-K cover gives a sense of what the market currently pays for the franchise: non-affiliates held $368.1 billion of common equity as of February 26, 2026, against 1,131,423,212 shares outstanding on October 2, 2026. Micron sits third of three in DRAM and third in HBM, so it is a price-taker in commodity parts and a challenger in the part that carries pricing power. That combination, not the aggregate memory market, is the arena worth modeling.

02

Competitive advantage

Does the company have a moat? A moderate one: capital keeps rivals out, the cycle keeps returns down

Micron's barrier to entry is real. Three companies supply nearly all of the world's DRAM, and building a leading-edge fab takes capital and process know-how that has accumulated over decades. What Micron has not shown is that the barrier protects returns through a full cycle. After-tax return on invested capital ran about 18% in fiscal 2022, about -11% in fiscal 2023, about 16% in fiscal 2025, then roughly 82% in fiscal 2026. Rating: Moderate. The concentration test passes cleanly; the persistence test does not pass yet.

Market position: a clear third, inside a band that has not moved

Public share estimates put Micron third in DRAM at roughly the low-to-mid 20s percent, behind Samsung in the low-to-mid 40s and SK hynix near 30%. In NAND it sits in the low double digits behind Samsung and SK hynix. In high-bandwidth memory it is the third supplier into volume, behind the same two rivals.

The ranking has held for five years, and the estimates keep Micron inside the same share band through a 50% revenue decline and through the current shortage. What the record supports is narrow but useful: no rival has displaced Micron, and Micron has not displaced either larger supplier. In a market where the low-cost producer sets the price, third place means accepting that price.

Market position: Fairly Strong. A share in the low-to-mid 20s is well short of the 40% level that would signal dominance, but the three-firm structure has survived a severe downcycle without a fourth supplier taking meaningful volume.

Profitability: an average built on one extraordinary year
Fiscal yearRevenue (USD bn)Gross marginNet marginROIC (est.)
202230.845.2%28.2%~18%
202315.5-9.1%-37.5%~-11%
202425.122.4%3.1%~2%
202537.439.8%22.8%~16%
2026133.280.7%63.8%~82%

The five-year average after-tax ROIC of about 21% clears the 15% threshold that usually separates advantaged businesses from the rest, and on that number alone Micron would look protected. The average misleads. Four of the five years average roughly 6%, the zone that normally indicates no advantage at all, and fiscal 2026 alone pulls the mean above the bar.

Micron's fiscal 2026 10-K shows where the money came from. Revenue rose to $133.19 billion from $37.38 billion, while cost of sales went from $22.5 billion to $25.7 billion, up 14%. Revenue rose 256%. Price and mix produced nearly all of the increase, and an 80.7% gross margin is what a shortage looks like on an income statement. Operating expenses fell to 6.1% of revenue from 13.4%, which is operating leverage on a large fixed cost base, though that base itself rose 63%.

That margin is not a durable feature. Micron earned 39.8% gross margin in fiscal 2025 and a negative gross margin in fiscal 2023.

Profitability: Moderate. The five-year average ROIC of about 21%, an estimate built from reported EBIT, tax and balance-sheet figures, clears the 15% threshold, but the range from about -11% to about 82% says industry supply sets the return, not a protected position.

Overall rating: Moderate until a down-cycle proves otherwise

Profitability is the binding constraint here, so the overall rating cannot sit above Moderate regardless of how stable the share looks. Micron has the barrier without the pricing power that a strong moat is supposed to deliver.

Rival-level returns are not available on a comparable basis: Samsung reports memory inside a much larger devices and foundry business, and SK hynix's segment disclosure does not separate HBM economics. The observable evidence is that Micron's gross margin traveled from negative to 80% in three years while its product lineup stayed broadly similar. That points to industry pricing as the dominant variable.

Two developments would change this judgment. The first is a down-leg: if gross margin holds above roughly 40% through the next supply correction, returns are being protected by something other than scarcity. The second is HBM share. If Micron holds a durable quarter or more of that market across two accelerator generations, the captive part of its business is larger than the record shows.

Where the advantage comes from and how long it can last: capital deters entrants, HBM lock-in lasts a generation

Micron's advantage is mostly supply-side, and supply-side advantages are the weakest kind: they keep new money out without giving the incumbent much hold over price. Two of the three sources below are cost and process. The third, HBM qualification, is the only one that creates customer captivity, and the captivity is short.

Capital intensity: it stops financial entrants, not strategic ones

Strength: Moderate. DRAM capacity has to be built before it can be sold, in fab shells and lithography tools that take two to three years to bring online, and it has to be paid for whether or not prices are good.

Micron's cash flow statement shows the shape of that commitment. In fiscal 2025, a decent year, capex was about $15.9 billion on $37.4 billion of revenue, implied by operating cash flow of $17.52 billion against free cash flow of $1.67 billion. In fiscal 2023, a year of negative gross margin, capex was still roughly $7.7 billion, about half of that year's revenue.

An entrant weighing that arithmetic must fund several years of losses before it has a competitive cost per bit. That is why the supplier list has been stable, and it is the source of whatever pricing discipline exists. The limit is who it deters. It deters a new publicly traded entrant. It does not deter Samsung, which funds memory from other divisions, and it does not deter a state-backed entrant whose capital does not depend on returns. Subsidies cut both ways: the US, Japanese and Indian incentives Micron has taken also lower the cost of entry for anyone who can negotiate a similar package.

This barrier has no expiry date, only a failure condition. It holds as long as leading-edge lithography keeps getting more expensive, which is a technology trend, and it fails if a government is willing to fund a loss-making entrant for a decade.

Process know-how: a lead measured in quarters

Strength: Moderate. Making DRAM at competitive cost is a yield problem. A new node produces fewer good die per wafer at first, yields improve with learning, and the supplier with the best yield at a given node can cut price and still earn a margin. Process recipes, design rules and test methods are protected by patents and trade secrets, which helps explain why the same three companies keep landing near the leading edge.

The edge leaks in two directions. Equipment is sold to everyone: ASML ships EUV to every DRAM maker that can pay, and Applied Materials, Lam Research and Tokyo Electron sell the same deposition and etch tools to each. And the price mechanism transfers the benefit. Micron's gross margin went negative in fiscal 2023, when its process technology was not materially behind. Cost leadership in this industry protects survival, not price.

HBM is where process know-how may carry further. Stacking dies with through-silicon vias, thinning wafers and bonding the stack to a logic base die is a less standardized problem than a DRAM shrink, so yield differences between suppliers may persist longer there. That is an inference from the structure of the process, not something Micron has quantified, and it would be worth at most one product generation at a time.

HBM qualification: captivity that expires with each accelerator generation

Strength: Moderate. HBM is designed into a specific accelerator, not bought from a catalog. Supplier and chip designer validate the stack together over several quarters, and once the part is qualified the designer is reluctant to requalify a second source mid-program, because a failed validation costs a product launch. The incumbent supplier gets pricing leverage for the life of the platform. This is the one part of Micron's business where customers negotiate allocation ahead of shipment instead of paying a spot price.

The capture is short. Every new accelerator generation reopens qualification, and buyers deliberately keep two or three HBM suppliers qualified to preserve their leverage. HBM is also a JEDEC standard: the specifications are public, and each generation has made parts from different suppliers closer to interchangeable. The lock-in protects a few quarters of revenue and some price, not a decade of returns. Micron being the third supplier into volume means it is winning sockets from a smaller base, which is where share gains would have to come from if they come at all.

Durability: Chinese capacity and standardization are the live threats

Two developments can shrink this moat.

The first is Chinese DRAM supply. CXMT is state-backed and has been adding commodity DRAM capacity, which lands on the DDR5 and LPDDR5 parts that remain Micron's volume base. Export controls on advanced equipment slow it, and the evidence so far does not establish whether it can yield at the leading edge. It does not need to. A credible supplier of commodity DRAM at scale would set the price for the products that carried most of Micron's bit volume before the shortage.

The second is a shift in where the value sits in the stack. If HBM keeps migrating toward custom logic base dies built by foundries, more of the differentiated content belongs to the foundry partner and less to the memory supplier. Separately, if AI inference moves toward cheaper memory tiers, GDDR or LPDDR can serve workloads that HBM serves today, which would cap HBM pricing without any new entrant appearing.

Micron's balance sheet is not a moat. Net cash of roughly $35 billion (cash of $38.36 billion against $3.28 billion of debt at the September 3, 2026 year end) lets it keep spending through a downturn, a genuine advantage in a capital-intensive industry, but it does not stop Samsung or CXMT from building. The AI memory story is demand, and demand is what produced the 80% gross margin. Neither is protection.

The DRAM barrier should hold for at least the next several years, since the leading edge has not become cheaper to enter and no fourth supplier has taken volume. The HBM advantage is smaller and shorter, and it is best treated as a per-generation asset. Whether the current returns are structural is answerable, and the answer arrives with gross margin in the next supply correction: near 40% and the moat deserves more than Moderate, back toward fiscal 2023 levels and the past five years were a cycle, not a franchise.

Generated by DeepVal AI. Data and models may contain errors or be outdated. For research only; not personalized investment advice.

Continue the comparison

Continue reading the full report

  1. Comparison and summaryHow Micron Technology Inc compares with its closest rivals, and our moat rating

Free to sign up · 20 research points · unlocking this report uses 10

20 points cover 2 full reports. No card needed.

Research reflects its generation date. Preview the first two modules.

More on Micron Technology Inc