Micron makes DRAM and NAND— memory chips that, for most of the company's history, were a boom-bust commodity: it lost $5.8bn as recently as FY2023.[6] Then AI turned memory scarce. In its February-2026 quarter Micron booked $23.9bn of revenue at a 74.4% gross margin — up 196% year over year — and around 26 May 2026 it briefly crossed a $1 trillion market cap.[1]·[44] The question is no longer whether Micron benefits from AI. It is whether this is a structural re-rating or the steepest peak of the same old cycle — and whether Micron can hold its place in HBM, where it is still a distant third.

Figures from Micron's FQ2-26 and FY2025 SEC results; market cap per trackers.[1]·[4]·[44]

Source: Micron results; FY revenue history per stockanalysis.com.[4]·[7]

The three questions this study is organised around

This is a compilation, not an argument. On each question the evidence genuinely cuts both ways; we lay out the strongest version of each side and leave the weighing to you.

Micron is an integrated device manufacturer (IDM) — it both designs and fabricates its own DRAM and NANDmemory — and the only US-headquartered firm in a global oligopoly otherwise made up of South Korea's Samsung and SK Hynix. It employs roughly 53,000 people and manufactures across the US, Japan, Taiwan and Singapore.[38]·[5]

What Micron actually makes

Micron sells two main kinds of memory. DRAM (dynamic random-access memory) is the fast, volatile working memory in PCs, phones, cars and — increasingly — AI servers, where it takes the form of HBM (high-bandwidth memory) stacked beside GPUs. NAND flash is the non-volatile storage in SSDs and devices. Unlike a fabless designer, Micron owns its fabs; unlike a foundry such as TSMC, it designs and brands its own products. That makes it a capital-intensive product company selling a largely standardised good — the defining tension of its business model.[38]·[48]

The Mehrotra era

Sanjay Mehrotra, who co-founded SanDisk and ran it until its 2016 sale, took over as CEO in 2017 and is now Chairman, President and CEO. Under him Micron survived the savage 2023 downturn — a $5.8bn loss year — pruned the failed 3D XPoint bet, and repositioned around AI memory, declaring Micron “the only U.S.-based memory manufacturer.”[39]·[42]·[5]

Selected timeline

Three companies — Samsung, SK Hynix and Micron — control roughly 93% of the global DRAM market; NAND is more fragmented. Memory has always been cyclical and commodity-like: prices swing with the supply/demand balance, and in the 2023 glut the whole industry lost money. AI is now the swing factor — it is projected to consume ~20% of global DRAM wafer capacity in 2026.[27]·[25]

Where Micron sits in the value chain

Unlike the logic world — where fabless designers (Apple, NVIDIA) hand their chips to a foundry (TSMC) — memory is still dominated by integrated device manufacturers that both design and fabricate. Micron buys tools from a concentrated equipment oligopoly upstream and sells to module makers, server OEMs and hyperscalers downstream. Its leverage comes from the one link it owns and few can: leading-edge memory fabrication.

HBM stacks Micron-made DRAM dies on a logic base die fabricated by TSMC — so even Micron now depends on a foundry for part of its flagship product.[12]

The market is enormous — and just repriced violently

Total DRAM industry revenue roughly doubled in a year, from $27.0bn in Q1 2025 to $53.6bn in Q4 2025, as contract prices jumped 45–55% quarter-on-quarter. NAND followed, with prices projected up 85–90% QoQ into early 2026.[21]·[24]

Source: TrendForce quarterly DRAM trackers.[23]·[22]·[21]

The cycle: why memory is feast or famine

Memory products are largely interchangeable, so they compete mostly on price and cost. When demand outruns supply, prices spike and margins balloon; the makers then invest heavily, supply overshoots, and prices collapse. The last bust was severe: Micron's FY2023 revenue halved to $15.5bn and it posted a $5.8bn net loss at a negative gross margin.[6] The bull claim is that AI has changed this — HBM is harder to make (1GB of HBM uses ~4× the wafer capacity of standard DRAM) and demand is pre-booked, so suppliers warn of shortages through 2027.[25]·[20]

Micron earns its margin on the gap between its cost per bit (driven by technology node, yield and scale) and the market price (driven by the cycle). That gap swung from a negative gross margin in FY2023 to 74.4% in early 2026 — same factory, wildly different economics — because price, not cost, moved.[6]·[1]

Micron FY2025 results and FQ1-26 call.[4]·[9]

How a memory maker actually makes money

Because DRAM and NAND are largely standardised, Micron can't charge a brand premium — it competes on cost per bit and, increasingly, on differentiated high-value products (HBM, high-capacity server DRAM, low-power data-center modules). Three levers drive cost: advancing to a denser process node (more bits per wafer), raising yield (fewer defective dies), and scale to spread a multi-billion-dollar fab. The fixed-cost base is huge, so when prices rise, almost all of it drops to the bottom line — and when they fall, the same leverage works in reverse.

The mix has shifted decisively to the data center

Micron reorganised its reporting around AI. In FQ2-26 the Cloud Memory and Core Data Center units together were the largest and fastest-growing slice, and every business unit earned a gross margin above 68% — even the historically thin mobile and embedded segments.[48]

Source: Micron FQ2-26 results (SEC 8-K). Total revenue $23.86bn.[48]

Source: Micron FQ2-26 results (SEC 8-K).[48]

The capital-intensity problem

The catch is that leadership must be continually re-bought. Micron spent $13.8bn of capex in FY2025 — roughly 37% of revenue — and is funding a multi-year, ~$200bn global buildout. In good years that investment is easily covered; in a downturn the bills keep coming while revenue collapses. Today demand so outstrips supply that Micron says it can meet only about half to two-thirdsof several key customers' needs — a seller's market that lets it pre-sell and price for the year.[4]·[9]

In DRAM, Micron is a clear No. 3 (~22–26% share) behind Samsung and SK Hynix; in HBM — the AI prize — it is third again, with SK Hynix the runaway leader. Micron's edge is technology and a fast HBM ramp; its vulnerability is scale and the risk of being squeezed by a resurgent Samsung above and Chinese entrants below.[21]·[13]·[26]

Five Forces: a concentrated structure under commodity pressure

Memory's consolidation into three players should make it a comfortable oligopoly — and at the cycle peak it looks like one. But the commodity nature of the product keeps rivalry high, suppliers hold real power, and the spectre of Chinese entrants and the next price war never fully lifts.

Positioning: who plays where

Mapping the field by product focus (commodity/legacy vs. leading-edge AI memory) against scale shows the structure clearly: SK Hynix and Samsung anchor the high-scale, AI-leading corner; Micron sits close on technology but below on scale; the NAND-only players and Chinese entrants occupy the lower-left. Hover a point for the basis.

The HBM scoreboard — where the money is

HBM is the component AI accelerators are starved for, so it is where the competitive battle matters most. Per Counterpoint, Q3 2025 HBM revenue share was roughly SK Hynix 57% / Samsung 22% / Micron 21%. On a bit-shipment basis Micron had climbed to ~21% from just ~4% a year earlier — a real gain — but a Samsung rebound pushed Micron back to third by revenue, and SK Hynix is set to take ~two-thirdsof NVIDIA's next-gen HBM4.[15]·[16]·[14]NVIDIA's Jensen Huang says all three are now qualified and in production for Vera Rubin.[47]

Source: Counterpoint Q3 2025 HBM revenue share (SK Hynix 57% / Samsung 22% / Micron 21%).[15] Bit-share gain from ~4% to ~21% per Mark Lapedus/Semiecosystem.[16]

The Chinese entrants

The long-term wildcard is China. CXMThas moved from legacy DRAM into DDR5-8000 and is already ~30% of China's smartphone LPDDR market; YMTC aims for 10%+ of global NAND capacity. Both are capped at the leading edge by export controls — but their cheap legacy output can flood the commodity tiers and pressure prices, exactly where Micron is most exposed.[26]·[27]·[28]

Micron's revealed strategy is to trade up the value chain — from commodity bits to differentiated AI memory where it can win premiums and share — while leaning on government industrial policy to fund a US/Japan capacity base. Its moat is real but rented: process leadership, scale and customer qualification must be re-won every node.[11]·[32]

Stated vs. revealed strategy

Micron saysit is “an essential AI enabler” built on “technology leadership” and its position as “the only U.S.-based memory manufacturer.”[5] What it actually doesis sharper than the slogans: it has pivoted capacity and R&D toward HBM and high-margin data-center products, pre-sold its scarce HBM for the year, killed the failed 3D XPoint experiment, and committed ~$200bn to a US buildout that it pairs explicitly with federal and state subsidies. The strategy is to be differentiated where it can and subsidised where it must.[8]·[32]

The technology claims — and the caveats

Micron points to a string of firsts: industry-first volume G9 NAND (3.6 GB/s, up to 73% denser), EUV-based 1-gamma DRAM, and HBM4in high-volume production at >11 Gb/s and >2.8 TB/s with a claimed >20% power-efficiency edge.[41]·[40]·[11] Two honest caveats: Micron was the last major maker to adopt EUV, and its headline HBM power-efficiency figure comes from its own internal calculator, not an independent benchmark.[40]·[11]

SWOT

What the moat is — and isn't

Micron's durable advantages are scale, deep process IP, capital, and the oligopoly structure of memory, reinforced by the stickiness of HBM customer qualification (once designed into an NVIDIA platform, a supplier is hard to displace mid-generation). What the moat is notis permanent: the product is still substantially a commodity, leadership is re-bought each node, and a sub-scale position means Micron earns the market's economics, not a monopolist's.[47]·[19]

Micron is a clear No. 3 in both DRAM (~22% in Q4 2025) and HBM (~21%), behind Samsung and SK Hynix. Its differentiators are being the only US-based makerand a fast HBM ramp — not size. The NAND-only peers (Kioxia, SanDisk) don't play in the AI-memory prize at all.[21]·[15]

DRAM revenue share (Q4 2025)

Source: TrendForce, Q4 2025 (total DRAM $53.58bn).[21]·[27]

HBM revenue share (Q3 2025)

Source: Counterpoint via Semiecosystem; bit-shipment basis differs.[15]·[16]

Side-by-side

DRAM/HBM shares are reported estimates that move quarter to quarter; live peer market caps are omitted except where cited (see Methodology).

In FY2025 Micron earned $37.4bn of revenue and $8.5bn of net income — a record — and the next two quarters were dramatically larger still, with FQ2-26 revenue of $23.9bn at a 74.4% gross margin. Two years earlier the same business lost $5.8bn. That is the amplitude of the memory cycle.[4]·[1]·[6]

Micron FY2025 and FQ2-26 SEC results.[4]·[1]·[2]

The revenue ramp is almost vertical

Quarterly revenue went from $8.05bn a year ago to $23.86bn in February 2026, and Micron guided the following quarter to $33.5bn at roughly 81% gross margin — with GAAP EPS of $18.90. First-half FY2026 revenue alone already exceeded all of FY2025.[1]·[2]

Source: Micron FQ1-26 and FQ2-26 results (SEC 8-K); FQ3-26 figure is company guidance.[3]·[1]·[2]

Cash, capital returns, and the buildout

Profitability has translated into cash: $17.5bn of operating cash flow in FY2025 and a growing net cash cushion ($16.7bn by FQ2-26), even while Micron spent $13.8bn of capex in FY2025 and is funding a ~$200bn multi-year expansion. Confident in the run, the board raised the dividend 30% to $0.15 per quarter.[4]·[1] The stock followed: MU rose roughly 10× in a year and briefly topped a $1tn market cap in late May 2026, with UBS setting a Street-high $1,625 price target.[44]

Three risks dominate: a cyclical reversal in memory prices; the loss of China as a market after the 2023 ban; and the execution and policy riskof a ~$200bn subsidy-backed buildout that won't produce US DRAM until 2027. The same US-champion status that brings CHIPS money also makes Micron a target in Beijing.[19]·[30]·[32]

China: a market largely lost

In May 2023, China's Cyberspace Administration barred Micron from sales to critical-information-infrastructureoperators, citing unspecified “serious cybersecurity issues.” China had been ~10.7%of Micron's FY2022 revenue; Micron estimated the ban put roughly $4bn of annual revenue at risk, and it has never recovered that business.[30]·[31]US export controls on advanced HBM to China since late 2024 cut the other direction, limiting what Micron can sell to Chinese AI customers.

The reshoring bet — asset and liability

Micron has tied itself tightly to US industrial policy. It secured up to $6.4bn in CHIPS funding, announced a ~$200bn US plan ($150bn manufacturing + $50bn R&D), and aims to make 40% of its DRAM in America. In Japan it is adding a ¥1.5tn (~$9.6bn) Hiroshima HBM fab with up to ¥500bn of METI subsidy, and the January-2026 Section-232 tariffs explicitly favour firms that build domestically.[32]·[34]·[35]The flip side: the economics lean on subsidies and policy that can change, and the first US DRAM output isn't due until the second half of 2027.[33]

The IP front: a cautionary tale

Micron's long fight with China's Fujian Jinhua over alleged DRAM trade-secret theft shows the limits of legal recourse. After a 2018 US indictment and Entity-List placement, the two firms reached a global settlement in December 2023, and in February 2024 a US judge acquitted Fujian Jinhua of the criminal charges (co-defendant UMC had earlier pleaded guilty and paid a $60m fine).[36] The episode underlined that protecting memory IP across the US–China divide is slow and uncertain.

The cyclical risk never sleeps

Underlying everything is the cycle. Today's 74% margins are a peak; the last trough cost Micron a $5.8bn loss, and skeptics warn the new fabs being built now — by all four major makers — set up the next glut in 2027–28.[6]·[19] Micron must fund its capex commitments through whatever that brings.

The bull and bear cases share the same facts and differ on one judgment: whether AI demand permanently raises the floor for memory or merely delays the next correction. Micron's technology and US position are genuine; its sub-scale HBM standing and commodity DNA are equally genuine. Where you land on the cycle decides almost everything.[25]·[19]

Bull · Base · Bear

The questions that decide it

Three things, watched over the next two years, will settle the debate. First, does HBM stay scarce? If HBM4/4E supply remains constrained and Micron defends its NVIDIA allocation, the bull case strengthens; if SK Hynix and a recovering Samsung add capacity faster than demand, Micron — the smallest of three — feels it first.[14]·[17]Second, when do the new fabs bite? Volume capacity from all four makers lands in 2027–28, the classic trigger for a price roll-over.[19] Third, does the geopolitical frame hold? CHIPS money, tariffs and the China standoff could each swing the economics either way.[32]·[35]

Method

Research proceeded by fan-out web search and direct fetching of primary and reputable secondary sources, working per section outward from Micron's own SEC filings and earnings releases to trade analysts, press and tertiary context. Because Micron is a US company, the work was done in English, but it deliberately reached for the home markets of Micron's rivals — SK Hynix's own newsroom, TrendForce and Counterpoint trackers, Japanese and Chinese fab/subsidy reporting — and ran disconfirming searches (memory cycle bust, HBM third place, China ban, overbuild, valuation peak) so the critical case is sourced as rigorously as the bullish one. Every URL cited was opened and read during the research run; no link appears unless it was fetched, and a link-checker validated each one while quotes were spot-checked against the source. Each claim was transcribed into a structured manifest tagging it with a tier, a confidence level and a stance. The load-bearing figures for Micron are its quarterly and full-year results — revenue, gross margin, segment mix and capex — taken straight from its SEC 8-K earnings exhibits.[1]·[3]·[4]

Frameworks used

The study applies the Pyramid Principle (answer-first, balanced) throughout, Porter's Five Forces and a 2×2 positioning map in the Competitive Landscape, peer comparables in Peer Comparison, a commodity-economics and segment-mix lens in the Business Model, a SWOT in Strategy, and bull/base/bear scenario analysis in the Forward View — each chosen because the disclosed data could support it even-handedly. Frameworks the data could not fill that way, such as McKinsey 7S or a full BCG portfolio, were skipped rather than padded out with guesswork.

Disclosed vs. estimated

A clear line separates what Micron discloses from what others estimate. The disclosed, high-confidence layer is Micron's own financials and operating metrics, drawn from its FY2025 and FQ1–FQ2 2026 results and SEC filings.[1]·[3]·[4]Comparable-basis figures — DRAM, NAND and HBM market-share percentages, competitor yields, the HBM allocation split for NVIDIA's Vera Rubin, Chinese-entrant capacity, and the live market cap — are directional analyst and press estimates (TrendForce, Counterpoint, Tom's Hardware), not Micron disclosures, and are flagged as such wherever they appear.[13]·[15]·[21]·[24]·[45]Where even a defensible estimate could not be cleanly re-sourced at the as-of date, a range is shown or the figure is excluded rather than guessed.

Neutrality & independence

This is a compilation, not an argument: every section pairs the case for and the case against, with supporting and critical evidence tagged by stance and frameworks applied even-handedly. It is an independent research artifact, not affiliated with, endorsed by, or sponsored by Micron Technology, Inc., provided for informational and educational purposes only and not investment advice; critical claims are attributed to named sources and framed as reporting or analysis, never as bald fact. It is point-in-time as of 6 June 2026, and errors are the author's — corrections welcome.

Tier 1 = primary / authoritative (Micron SEC filings and press releases, earnings releases, official company pages, US government documents). Tier 2 = reputable secondary (Reuters, CNBC, TrendForce, Tom's Hardware, The Register, Fortune, Counterpoint analysts). Tier 3 = tertiary / context (encyclopedias, market-cap trackers, aggregators) — used for colour, not load-bearing facts.