14 September 2026 · Built from Lam's SEC filings FY2016–FY2026 and roughly sixty earnings-call, investor-day and conference transcripts through 10 September 2026, plus independent evidence from SEMI billings statistics, BIS and Federal Register rulemaking, the Congressional Research Service, and the primary filings of ASML, Applied Materials, Tokyo Electron, KLA, TSMC, Micron, SK hynix, Naura and AMEC. This is not a valuation and not a recommendation.
A chamber is sold into a new fab as systems revenue — $14.89bn in FY2026. To be sold at all it must first be written into the customer's production recipe as the process of record, a qualification no rival can displace without requalifying the whole process and risking yield. Once installed it consumes spares, requires service, accepts upgrade kits at each node transition, and is eventually resold into mature-node production as a refurbished Reliant tool. Those four streams produced $8.35bn in FY2026 — 35.9% of revenue — from an installed base that passed 100,000 chambers in the December 2025 quarter.
A modern chip is built by depositing a film a few atoms thick across a silicon wafer, cutting a pattern into that film, cleaning away the residue, and repeating several hundred times. Lam sells the machines that do the cutting, the depositing and the cleaning. It does not sell the lithography scanner that projects the pattern — that is ASML — nor the inspection tools that check the result, which is KLA. It sells the steps in between, and describes itself as "a global supplier of innovative wafer fabrication equipment and services" with leadership claimed in "deposition, etch, and clean markets" (FY2026 10-K).
The products map onto what they do to the wafer. In deposition, ALTUS lays down tungsten and increasingly molybdenum; SABRE plates copper interconnect and, in its 3D form, builds the copper pillars and redistribution layers that hold an advanced package together; Striker and VECTOR deposit dielectrics. In etch, Akara handles the advanced conductor etch that gate-all-around transistors demand, Vantex cuts the extreme high-aspect-ratio holes through a 3D NAND stack, and Syndion etches the deep vias used in memory stacking. In clean, Coronus, DV-Prime and EOS remove what the other steps leave behind.
Lam's own disclosure is built around chamber counts rather than tool counts or customers, and the economics follow that unit exactly. A chamber is sold once, at the moment a customer adds capacity. It is then monetised continuously for as long as it runs — which, on Lam's account, is effectively forever. The important consequence is that support revenue compounds with the accumulated installed base rather than with the current year's capital cycle, and management explicitly targets support revenue growing faster than the chamber count, a target it reported hitting in every quarter of FY2026.
The second consequence is that an upgrade is economically different from a new tool sale. An upgrade kit converts existing customer capacity to a newer node in a fraction of the time and cost of buying and installing a new machine — which is why Lam says it captures "more than two-thirds of the available" served market in a NAND upgrade. The customer is buying capability rather than capacity, and Lam's entrenchment is what makes the sale possible at all.
One thing that's true is Lam tools rarely get retired… Our tools just don't find themselves obsoleted. And that's because of the power of upgrades… They continue to generate parts and service business for us for decades and decades.
Who pays is a very short list. Lam names Micron, Samsung, SK hynix and TSMC as its most significant customers, and concentration deepened in FY2026: four customers each exceeded 10% of revenue, at roughly 16%, 15%, 12% and 12%, against two customers in FY2025 and one in FY2024 (FY2026 10-K). Lam reports a single business segment, so there is no segment profit disclosure to work with; revenue is disaggregated only as systems versus support, and by geography.
Semiconductor equipment exists because chipmakers cannot build their own tools. Removing material selectively at atomic scale, or growing a uniform film inside a hole a thousand times deeper than it is wide, is a discipline no chipmaker has found it economic to internalise. Independent measurement comes from SEMI, which publishes actual billings rather than forecasts: global equipment billings were $117.1bn in 2024 and $135.1bn in 2025, up 15%. Wafer fab equipment is the front-end subset; management's working figure was roughly $110bn for calendar 2025, raised three times during FY2026 to "the low $150 billion range" for calendar 2026.
The market is not one market but a set of adjacent near-monopolies. ASML supplies EUV lithography with no alternative. KLA dominates process control. Tokyo Electron leads the coater-developer track. Applied Materials has the broadest portfolio. Lam competes principally with Applied Materials and Tokyo Electron in etch and deposition, and increasingly with Naura and AMEC inside China. The structure is durable because changing supplier for one step forces requalification of everything downstream.
| Company | Fiscal period | Revenue | Gross margin | Op. margin | China % rev. |
|---|---|---|---|---|---|
| KLA | FY25 (Jun-25) | $12.16bn | 60.9% | 38.2% | 33% |
| ASML | CY2025 | €32.7bn | 52.8% | 34.6% | n/d |
| Lam Research | FY25 (Jun-25) | $18.44bn | 48.7% | 32.0% | 34% |
| Applied Materials | FY25 (Oct-25) | $28.37bn | 48.7% | 29.2% | 30% |
| Tokyo Electron | FY26 (Mar-26) | ¥2,443.5bn | 45.3% | 25.6% | 34.1% |
| Lam Research | FY26 (Jun-26) | $23.23bn | 50.5% | 35.3% | 33.8% |
Read across the FY2025 row. Lam's gross margin matches Applied Materials and beats Tokyo Electron, but it converts that into a higher operating margin than either, because it spends less of its revenue on R&D and SG&A than its two direct rivals while still spending $2.1bn. KLA's higher margins reflect a different product — metrology carries more software and less metal — not a better-run version of the same business. The instructive comparison is Tokyo Electron, whose FY2026 operating margin fell 310 basis points to 25.6% on record revenue, attributed by the company to "soaring costs of parts and materials." Lam's margin rose over the same window. Cost structure, not the cycle, separates the two.
Equipment is not the most profitable place in this value chain at the current point in the cycle. TSMC earned a 59.9% gross and 50.8% operating margin in calendar 2025, and SK hynix a 49% operating margin — both ahead of every equipment maker. The equipment firms cluster in a tighter 26–38% operating band. What they give up in peak profitability they gain in stability: Micron's 26.1% operating margin in FY2025 sits below the equipment group despite being a recovery year, and memory's ability to swing from that to a loss is why equipment makers, who sell to all of them, carry less end-market risk than any single chipmaker.
Two, and only two. The first is sustained R&D — all five majors spend between 11.2% and 14.4% of revenue on it, through downturns as well as upturns. An entrant would need to match a figure in the billions for several years before earning a first qualification, against incumbents who have already amortised the same work across a large installed base. The second is process-of-record qualification, which converts that spending into a defensible position: once a chamber is in a customer's recipe, replacing it risks yield on a line worth billions a year. The proof that this binds is Lam's own installed base — 100,000+ chambers accumulated over decades, generating $8.35bn of support revenue that exists precisely because tools are not removed. Capital intensity, by contrast, is a weak barrier: Lam's capex was 4.2% of revenue.
Lam's account of its own industry is largely consistent with independent data, but two load-bearing claims cannot be verified from primary sources. That etch and deposition gain share of WFE as devices become three-dimensional is partially supported: Tokyo Electron's own disclosure shows etch is its single largest equipment category at 35% of new-equipment sales, and SEMI's application data shows NAND equipment spending growing faster than the total — but the specific magnitudes Lam cites (served market per wafer rising 1.7–2.0× across the DRAM, NAND and foundry transitions) are company figures with no independent corroboration. That Lam leads in etch and deposition is unverified: neither Applied Materials nor Tokyo Electron discloses those revenues separately, and no neutral share data was obtainable from primary sources.
The China claim is where outside evidence is most useful, because it partially contradicts the comfortable reading. Naura grew calendar 2025 revenue 30.9% to RMB 39.35bn and AMEC 36.6% to RMB 12.38bn — both roughly twice the growth rate of global billings. Lam's explanation is that these gains sit entirely in accounts it is legally barred from serving: "there's a whole bunch of customers that used to be very big, important customers for us that we are prohibited from selling to today… That's where the Chinese guys are doing really well" (Doug Bettinger, Morgan Stanley conference, 3 March 2026). Plausible, and untestable from public data. What is observable is that Naura's net profit fell 1.8% while revenue grew 30.9% — the signature of a firm buying share on price. A competitor that behaves that way in a protected market can behave that way in an open one if the rules change.
What kind of company wins here is not ambiguous: the firm that holds an unassailable position in a specific process step, converts it into an installed base large enough to fund the next node's research, and sustains R&D through the trough when a weaker rival cuts. Lam is that kind of company. The qualification is that its position sits in etch and deposition, contested by two rivals of comparable scale, rather than in lithography, contested by no one.
Lam's revenue is the product of two numbers: how much the industry spends on wafer fab equipment, and what share of that spend falls inside Lam's served markets and is won by Lam. FY2026 revenue of $23.23bn was 26.0% above FY2025, with systems up 29.5% and support up 20.2%.
| Period | Reported growth | Acquired contribution | Organic growth |
|---|---|---|---|
| FY2026 vs FY2025 | +26.0% | nil — no deals FY24–FY26 | +26.0% |
| FY2016 → FY2026 CAGR | ~14.7% | ~$291m total consideration, 11 years | ~14.7% |
Ranked most to least impactful, these are the levers — each tagged for what kind of driver it is.
Management's calendar-2026 view rose three times during FY2026 — $135bn in January, $140bn in April, "the low $150 billion range" in July — against roughly $110bn in calendar 2025. This is the single largest determinant of any given year's revenue and the one Lam controls least. It is the reason the FY2026 figures look the way they do.
Lam's served share of WFE was in the low 30s in 2024, targeted at the high 30s at the 2025 Investor Day, and reached "north of 36%" by September 2026 — which the CFO described as "accelerated… more quickly, perhaps, than we had thought back then." Ship share of WFE grew "well over 1 percentage point year-on-year" in FY2026. This lever compounds independently of the cycle, which makes it the most important thing to track.
Moving from FinFET to gate-all-around makes the transistor genuinely three-dimensional, raising both the number of atomic-layer deposition and selective-etch steps and the difficulty of each — contact etch selectivity requirements rise from roughly 5:1 to 100:1. Lam quantifies the result as roughly "$1 billion in incremental Lam SAM for every 100,000 wafer starts per month of capacity," and a doubling of served market per foundry-logic wafer. Served by the Akara platform, whose installed base management says "has doubled every year" since launch.
As transistor density scaling slows, performance increasingly comes from stacking chiplets and memory inside one package — which requires the copper pillars, redistribution layers and through-silicon vias that SABRE 3D and Syndion build. Shipments exceeded $1bn in calendar 2024. Expected calendar-2026 growth was raised twice inside the year, from "more than 40%" in January to "greater than 70% year-on-year" in July.
Roughly two-thirds of industry NAND bit capacity sat below 200 layers at the start of 2025, and converting it requires an estimated "over $40 billion" of customer spending on tool types that do not exist below 200 layers. Lam is the process tool of record at every customer running above 200 layers. The timing changed materially during FY2026: what was a three-to-five-year conversion is now expected to see "the majority of spending occurring before the end of calendar year 2027." That pulls revenue forward and concentrates it.
Support revenue grew 20.2% to $8.35bn in FY2026 while the chamber count grew more slowly. Three of its four components — spares, service and upgrades — are driven by utilisation and node transitions rather than new capacity; the fourth, Reliant, tracks mature-node investment and is the most cyclical. Management describes operating margin here as "above the corporate average" without quantifying it, and targets support revenue at 1.5× its 2024 level by 2028.
Molybdenum metallisation replaces both the tungsten fill tool and the titanium-nitride barrier tool as tungsten runs out of resistance headroom, which Lam says makes the opportunity "two times what it was during the Tungsten era"; it targets over $2bn of shipments across three to five years. Dry photoresist replaces solvent-based coat and develop for EUV patterning, with a $1.5bn cumulative five-year revenue target and one named memory production win. Both are targets rather than results, and neither has been quantified as revenue since.
China was 34% of FY2026 revenue. Management guided a roughly $600m calendar-2026 impact from the affiliates rule when it was announced, and said the restriction would take China below 30% of revenue for the calendar year. Within China, global multinationals are growing while "the domestic customer base declined" — a mix shift that policy relief alone would not reverse.
The demand mechanism underneath all of this is AI infrastructure spending, which management translates into equipment demand at a stated ratio of "roughly $8 billion of WFE spending for every $100 billion in incremental data center investment," revised upward by "a billion or two" during FY2026. It is a useful rule of thumb and an explicit acknowledgement that Lam's revenue is a derivative of a capital-spending decision taken several layers up the chain.
Gross margin has risen from 44.5% in FY2016 to 50.5% in FY2026, crossing 50% for the first time in the company's history. Six hundred basis points on $23bn of revenue is worth more than any single growth lever, so the mechanism matters — and management separates it honestly into three parts.
The first is the manufacturing footprint. The Malaysia factory opened in 2021 and is now "the largest factory in our global network," with a second Malaysian facility ramping in the second half of calendar 2026. Moving incremental volume to Asian factories close to the customers cuts both the production cost base and freight spending. The second is operating leverage: R&D fell from 11.4% to 10.2% of revenue in FY2026 and SG&A from 5.3% to 4.9%, even as R&D spending rose $279m in absolute terms. The third is customer mix — which management explicitly flags as the impermanent one: "Admittedly, we've had some favorable customer mix, but a good portion and a sustainable portion of that growth in gross margin has been how we've pivoted the factory footprint." That candour has been borne out; FY2026 guidance repeatedly cited customer mix, alongside aluminium and steel tariffs, as a coming headwind.
Cash conversion is good but not perfect. FY2026 operating cash flow of $5.86bn came in below net income of $7.27bn, and actually fell slightly from FY2025's $6.17bn despite net income rising 36% — a sharply accelerating revenue line absorbs cash into receivables and inventory. Free cash flow was approximately $4.89bn after $966m of capex.
| FY2016 | FY2021 | FY2024 | FY2026 | |
|---|---|---|---|---|
| Revenue | $5.89bn | $14.63bn | $14.91bn | $23.23bn |
| Gross margin | 44.5% | 46.5% | 47.3% | 50.5% |
| Operating margin | 18.3% | 30.6% | 28.6% | 35.3% |
| Diluted EPS | $0.52 | $2.69 | $2.90 | $5.76 |
| Free cash flow | $1.17bn | $3.24bn | $4.26bn | $4.89bn |
| Diluted shares | 1,752m | 1,453m | 1,320m | 1,261m |
Over the eleven fiscal years FY2016–FY2026 Lam generated approximately $35.3bn of free cash flow and allocated it in an order that has not changed in a decade.
| Use of cash | FY16–FY26 | What the ranking reveals |
|---|---|---|
| Share repurchases | $27.47bn | The default use of cash. Diluted shares fell 28.0%, from 1,752m to 1,261m post-split. |
| Dividends | $7.97bn | Raised every year of the period; the quarterly rate reached $0.26 in FY2026. |
| Capital expenditure | $4.63bn | 4.2% of FY2026 revenue. Asset-light relative to its own customers. |
| Acquisitions | ~$0.29bn | Three deals in eleven years. Growth is built, not bought. |
| Net debt reduction | ~$0.60bn | Debt raised and retired opportunistically, not on a deleveraging plan. |
Buybacks and dividends together came to $35.44bn against $35.26bn of cumulative free cash flow — approximately 100% returned, consistent with the 75–100% policy set at the 2020 Investor Day. FY2026 alone saw $3.85bn of repurchases at an average $325.14 and $1.27bn of dividends, with $4.04bn remaining under the May 2024 authorisation. This is a company that decided its own shares are the best available use of surplus capital and has acted on that view through two downturns.
The one large capital decision that did not happen is worth recording. Lam agreed to acquire KLA-Tencor in October 2015 and terminated in October 2016 "after careful review of recent antitrust agency feedback." No break fee was paid, but the bond indenture forced a mandatory redemption of $1.6bn of senior notes at 101, producing a $36m loss on extinguishment in FY2017, with $51m of transaction costs expensed in FY2016. Lam has not attempted a transformational acquisition since.
Management has set out a financial model twice. The 2025 Investor Day targeted calendar-2028 revenue of $25–27bn, gross margin approaching 50%, operating margin of 34–35% and EPS of $6–7. FY2026 has already delivered 50.5% gross margin and 35.3% operating margin — at or above the 2028 target, two years early — while revenue and EPS remain below it. The margin targets were beaten because a structurally improved cost base arrived alongside a cyclical revenue surge. The prior 2020 model missed in the opposite direction: FY2024 revenue landed inside its $14.5–15.5bn range but its 32–33% operating margin target was missed at 28.6%, because the target year coincided with a trough.
| End market | FY2026 | FY2025 | Region | FY2026 | FY2025 |
|---|---|---|---|---|---|
| Foundry | 54% | 45% | China | 34% | 34% |
| Memory | 39% | 42% | Taiwan | 22% | 19% |
| Logic / IDM | 7% | 13% | Korea | 19% | 22% |
| Japan / US / SE Asia / Europe | 25% | 25% |
This is a high-cyclicality business with a well-documented record. Revenue fell 12.9% from FY2018 to FY2019 with operating margin compressing from 29.0% to 25.5%, and 14.5% from FY2023 to FY2024 with operating margin falling to 28.6%. Industry-wide, SEMI recorded global equipment sales down 7% to $59.8bn in 2019 and fab equipment spending down 15% to $84bn in 2023. Two things follow: peak-to-trough declines of 13–15% are what this business has actually delivered, and gross margin held up far better than revenue in the last trough — 47.3% in FY2024 against 44.6% in FY2023.
At or very near a peak, on every measure. FY2026 revenue is 33.3% above the prior cyclical peak of FY2023 and 110% above the FY2018 peak. Gross margin is 585 basis points above FY2023, operating margin 560 basis points above. These are not merely good numbers — they are the best numbers in the company's history, all arriving in the same year.
Management's own description is unusually unguarded: "Things are sold out, right? The industry is fundamentally sold out. The industry is constrained right now by clean room availability… the intensity of those conversations, the conviction of those conversations, I have never seen stronger in my time in the industry" (Doug Bettinger, Goldman Sachs conference, 10 September 2026). Fab utilisation is described as "basically at 100%," lead times are "stretching out," and customer discussions run "out into 2027 and beyond." Asked directly about overbuilding, his answer was candid in both directions: "This industry has always had a level of a cycle to it. That's always in the back of everybody's mind right now, but it's pretty far in the back… I'm more concerned for the company about missing the upside."
The downside mechanism this business would actually feel is not a demand shock but a capacity catch-up. The current upcycle is held up by a physical constraint — clean-room floor space — rather than by demand, and management described 2026's upward WFE revisions as customers who "squeezed out a little bit of extra space" and "resolved bottleneck tools." Eight to ten new tier-one fabs are described as coming online through the end of 2027. When that space arrives, the constraint forcing customers to bid for scarce tool slots is released, and orders normalise to underlying demand rather than to pent-up backlog. That transition, not an AI spending collapse, is the realistic mechanism for the next downturn — and it has a date attached to it.
| Indicator | Why it matters | Where published |
|---|---|---|
| Global equipment billings and fab spending | The only independent read on whether industry spending is still rising. Lam's own WFE figure is a forecast, not a measurement. | SEMI, monthly and quarterly (semi.org) |
| China as % of Lam revenue | The single most policy-sensitive line — it fell from 34% to 26% of revenue within FY2026. | Quarterly earnings release and 10-Q |
| Support revenue vs. chamber count | The direct test of whether the installed-base compounding thesis is working. | Quarterly earnings call |
| Served-market share of WFE | Currently ~36% against a high-30s target. The lever that works independently of the cycle. | Quarterly calls and investor conferences |
| Affiliates-rule suspension expiry | The one-year suspension lapses around November 2026. | BIS rulemaking, Federal Register |
| Memory-maker capex guidance | Leading indicator for the 39% of revenue that is memory-driven. | Samsung, SK hynix and Micron results releases |
These are the risks cyclicality does not already capture. The cycle is covered above; this is what compounds with it.
Genuine gaps, not omissions. Each would have to be resolved before a thesis could rest on unit economics rather than on the aggregate.