Technology
High-bandwidth memory, the DIMMs in an AI server and the memory in a phone all come off the same kind of line at the same three companies. That shared origin is why AI demand moves the price of memory in unrelated products, and why the choice of what to build is a capital allocation decision rather than a technical one.
In one sentence
DRAM manufacturing is the fabrication of dynamic random-access memory dies on silicon wafers, a capital-intensive process operated at scale by three companies and shared across every product category that uses DRAM.
A DRAM cell is a transistor and a capacitor, and the difficulty is that the capacitor has to hold enough charge to be readable while occupying an ever-smaller footprint. The answer has been to build it vertically — deep, narrow trenches with very high aspect ratios — which is why memory drives demand for the hardest etch and deposition processes in the industry.
The industry is three companies with fungible capacity. A wafer started as stacked memory is a wafer not started as a commodity module, and stacked memory consumes far more silicon area per gigabyte because of the logic base die and the yield loss in stacking. Allocating capacity toward it therefore tightens everything else, which is the mechanism behind memory price cycles that appear to have nothing to do with their apparent cause.
Three suppliers, a nearly interchangeable product, and capacity that arrives in large increments after a two-year build. When demand exceeds supply, price rises steeply because buyers have no alternative; when capacity lands into a surplus, price collapses because there is no product differentiation to defend. Memory is the most cyclical part of the semiconductor industry for structural reasons.
Shrinking the cell in two dimensions has slowed, so capacitor structures have gone deeper and process complexity has risen faster than density. That has made memory a major consumer of high-aspect-ratio etch, and it is why the equipment intensity of a DRAM line has climbed even where the node label has barely moved.
A stack of dies plus a logic base die, joined with through-silicon vias and yielded as a unit, consumes several times the wafer area per usable gigabyte that a conventional part does. That is why a modest share of bit demand can absorb a large share of wafer capacity, and why the two markets tighten together.
4 of these are marked as a chokepoint: a handful of qualified suppliers, a multi-year lead time, or a single geography.
Vertical capacitor structures are the hardest etch problem in volume manufacturing, and memory is its largest customer.
Deposition and etchMost DRAM layers are still printed on immersion tools, several passes each. Scanner deliveries, not cleanroom space, set how quickly a new memory fab reaches volume.
DUV immersionThe leading nodes have moved their tightest layers to EUV to remove multi-patterning steps. A maker without EUV access pays for it in cycle time and wafer cost on every die.
EUV lithographyEach patterning pass consumes qualified resist and the developers and strippers around it, from a short list of mostly Japanese suppliers. A qualification change is a months-long exercise per layer.
Resists and chemicalsOther pages in this map that name DRAM manufacturing as something they cannot do without.
What each company supplies at this step, and — where a public figure exists — its share of this specific market — with what that share measures, the period it covers and who published it. Some rows also show the company’s own reported revenue for the segment covering this step, which is a different thing: it says how much this business matters to that company, not how much of the market it holds. Not a ranking and not a recommendation.
Manufactures DRAM across commodity, server and stacked memory products.
17% high-bandwidth memory revenueQ2 2025 · Counterpoint Research
Manufactures DRAM and is the largest supplier of stacked memory.
62% high-bandwidth memory revenueQ2 2025 · Counterpoint Research
The US-headquartered DRAM manufacturer, supplying commodity, server and stacked memory.
21% high-bandwidth memory revenueQ2 2025 · Counterpoint Research
DRAM Products segment — 77.1% of Micron Technology’s reported revenue ($28.6B)FY2025 · company filings · not a market share
Supplies the etch and deposition equipment memory scaling depends on.
China's only DRAM manufacturer at scale, built to serve the domestic market that export controls closed off.
Supplies commodity and specialty DRAM outside the three leading-edge manufacturers.
Supplies the lithography systems that set how far the memory cell can be shrunk in each generation.
Supplies the deposition and etch tools that build the capacitor structures a DRAM cell is made of.
Supplies the coater/developer tracks and etch systems that sit between the lithography steps.
Supplies the inspection and metrology that establishes whether a shrink is yielding, which is what makes a node economic or not.
China's largest equipment maker, and the domestic tool base the capacity closed off by export controls has to be built on.
Supplies the atomic-layer deposition that builds the capacitor films a modern DRAM cell depends on.
Supplies the specialty and niche DRAM the three leading manufacturers no longer make in these volumes.
Supplies the silicon wafers underneath all of it; a shortage here caps every manufacturer at once.
Whether stacked memory keeps absorbing a rising share of wafer capacity.
Whether the three-supplier structure holds or a credible fourth entrant emerges.
Whether vertical scaling continues to raise equipment intensity per wafer.
Each page explains one technology in plain language, states what it depends on, and names companies by what they supply at that step. Company roles are described qualitatively and deliberately carry no market shares, revenue figures or rankings — those change faster than an explainer can, and a stale number is worse than none. Ticker links point at company pages on this site and are provided for reference only.
Nothing here is investment advice, a recommendation, or a forecast. A company named on a page about a technology is not thereby a good investment, and the chokepoints described are structural facts about supply chains rather than predictions about prices. Technology moves; where a page describes something as unresolved or in development, that was true when it was written.
Plutux is not an investment adviser. Market data and AI-generated analysis are for information and education only, not investment advice. Disclaimer