What this covers

This is a single-host walk through ASML by Tom, tracing the Dutch lithography maker from a Philips spin-out, kept alive through the severe 1980s industry cycles by cash from its backers, to the supplier of essentially all the world's most advanced lithography equipment. The technical thread runs through photolithography, the step that has gated Moore's Law since the 1970s. It covers why 193-nanometer deep ultraviolet light stopped reproducing shrinking patterns faithfully, why the industry chose extreme ultraviolet lithography in the 1990s, and why EUV reached production in 2019, roughly 13 to 14 years late. Physical details include a laser striking tin droplets twice, 50,000 times a second, and machines the size of a double-decker bus.

The business discussion centers on ASML as an architect and integrator: about 80 percent of cost of goods sold is bought-in components. It covers the modular design that derisks upgrades, the 2013 purchase of Cymer, and the 2012 deal in which Intel, Samsung, and TSMC took 23 percent and put roughly EUR 1.4 billion into R&D. Nikon and Canon abandoned EUV, and the research itself began in a US consortium. Field upgrades of an installed base, customer concentration, free cash flow conversion, and R&D near 15 to 16 percent of revenue round out the economics.

Several points are contested or surprising. Tom argues ASML's real competitor is Moore's Law itself, that sole-source status makes revenue less cyclical than the industry's, and that luck played a part in its early history. Other threads include Martin van den Brink's repeatedly underestimated forecasts, process node names that are now largely marketing, and NAND makers' shift to 3D stacking while EUV stalled.

Sharpest takeaway

ASML's dominance was built through patient, modular, collaborative industrial strategy rather than an initial advantage, and its sole-source position in EUV lithography, backed by customer co-investment and non-gouging pricing, makes it very hard to replicate.

  • ASML began as a weak Philips spin-out with no revenue or credible product and survived its first decade on cash injections
  • Only ASML had the funds and support to push EUV to market after Nikon and Canon gave up
  • ASML shares productivity gains with customers and outsources most components, which sustains the co-dependent ecosystem

The argument · threads9 threads · 54 claims
0.66

ASML's EUV position is unmatched, and no rival can catch up in the near term.

5 pointscentrality 5/5
  • ASML's share of wafer fabrication equipment spending is roughly 20 to 25 percent of the cost of building a new fab, higher for logic chips such as those made by Intel and lower in parts of the memory market.

    “ASML tends to account for somewhere between 20% and 25% of that. It varies according to the type of semiconductor.”

  • Having studied many technological advantages, Tom believes ASML's EUV position is more significant than any he has come across, given more than 25 years of development and over EUR 10 billion of R&D spending, and that nobody could catch up even in a decade.

    “So of all the competitive advantages, of all the technological advantages that I've ever come across, I can't think of one that's more significant than ASML's.”

  • Nikon had an EUV prototype around 2007, but the cost and technical difficulty led first Canon and then Nikon to give up on EUV, leaving ASML as the only player with the funds, resources, and support to bring it to market.

    “ultimately first Canon and then Nikon gave up on it.”

  • The EUV research was developed by a US consortium funded through the Department of Energy and DARPA, partly to keep the technology from Japanese competitors, and because no domestic US lithography maker could carry it forward, the consortium invited ASML to join.

    “But they didn't want to give it to the Japanese, and they didn't have a domestic lithography maker that was going to be capable of taking it forward.”

  • ASML has not been acquired because the semiconductor industry is carved into specialized niches with little vertical integration, and a Samsung or TSMC takeover attempt a decade ago would likely have prevented EUV development, since the other chipmakers would not have co-funded R&D under a rival owner.

    “They certainly wouldn't have had cooperation from the other major semiconductor manufacturers in funding the R&D that was necessary to make EUV work.”

0.54

Photolithography gates Moore's Law, so lithography stalls halt transistor density scaling.

6 pointscentrality 4/5
  • EUV was originally planned for introduction in the mid-2000s, but it was still not ready by 2010, and the first machines only came into use in 2019, roughly 13 to 14 years late.

    “finally, in 2019, it will take 13, 14 years after it was supposed to be delivered, these first EUV machines came out.”

  • Process node names such as 5-nanometer and 3-nanometer no longer correspond to the physical size of any feature on the chip and are largely marketing terms, though the industry still keeps squeezing more and smaller transistors onto silicon.

    “They're slightly marketing terms.”

  • Photolithography works by coating a silicon wafer with light-sensitive photoresist chemicals and projecting a circuit pattern through a mask with light and lenses, similar to a cinema projector but shrinking the image instead of enlarging it.

    “it's actually projecting a light to create a pattern on the surface of the silicon chip.”

  • As transistor patterns shrank, light wave behavior made deep ultraviolet light at 193 nanometers inadequate, because the wave-based light no longer reproduced the mask pattern faithfully, so the industry chose extreme ultraviolet as the next-generation fix in the 1990s.

    “as the patterns you're trying to draw get smaller, clearly the fact that light goes through a pattern then continues on in a wave becomes a problem.”

  • Photolithography has been the key gating technology for Moore's Law since the 1970s, so whenever photolithography stalls, the doubling of transistor density stalls too, as happened in the late 1970s and about ten years ago during the EUV struggle.

    “whenever photolithography is stalled, Moore's Law has stalled.”

  • When EUV was not ready around 2012-2013, NAND memory makers moved to three-dimensional stacking, building up layers through etch and deposition instead of scaling two-dimensionally, which sharply reduced lithography's share of capital spending in that segment.

    “So the share of lithography's CapEx within that part of the market fell quite precipitously as a result”

0.52

A few deep-pocketed customers and field upgrades lock ASML into its EUV dominance.

5 pointscentrality 4/5
  • Only a small number of very large, deep-pocketed companies can make leading-edge semiconductors, so they are the only customers able to afford ASML equipment and to use it effectively.

    “There are relatively few companies left in the industry that are capable of making leading-edge semiconductors.”

  • About 90 percent of lithography machines sold in the last 30 years are still operating in fabs, and ASML upgrades them in the field, so a machine installed a decade ago can gain substantial productivity, which keeps customers buying from ASML because the machine improves over decades.

    “90% of all of the lithography machines they've sold within the last 30 years are still in operation”

  • ASML's relationship with customers and suppliers is codependent and quid-pro-quo: customers fund and buy at returns that cover ASML's R&D risk, while ASML avoids gouging, and the industry's collaborative model depends on honesty and shared risk and reward.

    “So you have this -- what looks from the outside quite a comfortable position in terms of having a relatively small number of customers who are in themselves and comparatively powerful, but they're really codependent on each other.”

  • Extreme ultraviolet lithography was first used in high-volume production in 2019, giving ASML essentially 100% of the world's most advanced lithography equipment and 90%-plus of the next generation as well.

    “ultimately that led in 2019 to the first use in high volume production of extreme ultraviolet lithography equipment.”

  • ASML sold only 345 photolithography machines last year, with the most expensive costing more than EUR 150 million each, so the revenue of about EUR 21 billion reflects low-volume, high-price equipment.

    “It only sold 345 photolithography machines last year. It's a mass-produced item, but it's a very low volume, very, very specialized item.”

0.44

ASML's leadership and position were forged through necessity, luck, partner funding, and acquisitions after a troubled spin-out.

14 pointscentrality 2/5
  • ASML overtook Nikon to become industry number one in 2002, nearly twenty years after its spin-out, a goal it had set itself at the time of the spin-out.

    “2002, they finally overtook Nikon as industry #1. That's a goal they set themselves on spin out, took nearly 20 years to get there.”

  • In 2012, ASML persuaded Intel, Samsung, and TSMC to co-invest in the company, taking a 23 percent share between them, and this brought roughly EUR 1.4 billion of R&D funding into ASML.

    “finally, in 2019, it will take 13, 14 years after it was supposed to be delivered, these first EUV machines came out.”

  • The EUV light source fires a laser at tin droplets smaller than a dust particle, striking each one twice (once to flatten it, once to vaporize it into plasma about 40 times hotter than the sun) about 50,000 times per second in a vacuum, a capability ASML had to acquire by buying light source provider Cymer in 2013.

    “you have to strike each dust particle twice. Once to flatten it, once to vaporize it, to turn it into plasma, which is 40 times hotter than the surface of the sun. And you have to do that 50,000 times a second.”

  • ASML was spun out of Philips in 1984 as a problem child with no revenue, no commercially credible product, and no offices, and the joke inside Philips was that its creation was a delayed layoff process for transferred workers.

    “the joke within Phillips when it was spun out was that ASML being created was essentially a sort of delayed layoff process for the workers that were being transferred into it”

  • ASML's first decade was mainly about survival, as it scraped through severe 1980s industry cycles and was kept alive by further cash injections from its backers Philips and ASM, while its competitors faltered and fell away.

    “They were kept alive by further cash injections by their backers Phillips and ASM.”

  • The biggest operational risk is that suppliers must keep up with ASML's technology; ASML bought light source maker Cymer in 2013 and took a stake in lens maker ZEISS because the capital needed for next-generation equipment exceeded what suppliers could finance alone.

    “that was just recognizing that the amount of capital investment required to bring the next generation of equipment to market was beyond the budget of ZEISS as an entirely standalone company.”

  • Customers often pay deposits up front for equipment, which has helped ASML fund heavy R&D and capacity build-out, since development machines are often the only units able to take upfront deposits.

    “customers pay a down payment. They pay for this equipment often there's only the development machines that are able to take deposits upfront.”

  • ASML's outsourced assembler model was born of necessity, because it lacked the money and time to vertically integrate like Nikon and Canon, and this forced a supplier partnership model that Tom believes is now a core strength.

    “it's a core strength that actually it's become a virtue, but it actually was borne out of necessity at the company's beginning.”

  • Luck played a role in ASML's early history because a major 1980s recession killed some competitors and starved others of R&D funding, letting ASML catch up while others were on their knees, and ASML's position as the only non-Japanese EUV partner was also favorable.

    “there was a massive recession and it killed a couple of their competitors.”

  • The Japanese entered photolithography by copying American projection machines and then rapidly iterating until their machines were better, and they had in-house lens expertise, which was a key challenge of the technology.

    “starting initially, really essentially by copying American machines that were on the market, but rapidly, iterating to the point that their machines were simply better.”

  • ASML's current CEO, who joined in 1999 and became CEO in 2013 after serving as CFO, kept money flowing into R&D during the post-2000 tech-bubble downturn, which Tom credits as central to ASML's success.

    “he ensured that the money kept going into R&D, so the company kept extending his advantage.”

  • Martin van den Brink joined Philips in 1983 as his first job, led the make-or-break EUV-precursor product development after his boss left within two years, became CTO in 1995, and has remained at ASML since, and Tom considers him close to a genius.

    “And in ‘95 he stepped up onto the management board and was appointed CTO.”

  • The competitive position of ASML was built over decades through persistence and investment, not handed to it at the start, making it an example that dominance is earned through long execution.

    “And often you think of the dominant players having sort of started with some massive competitive advantage.”

  • ASML's early survival depended on backers, and its cash generation reflects in part customers paying deposits; both conditions show that financing structure, not just product quality, decided outcomes.

    “customers pay a down payment.”

0.38

ASML restrains its pricing and power because customers could otherwise pursue alternative technologies.

5 pointscentrality 3/5
  • Although ASML could theoretically double prices given its monopoly, it does not price gouge because that would give customers a strong incentive to pursue alternative technologies and providers, so it prices on the productivity improvements delivered and aims to split the benefit roughly 50-50 with customers.

    “In theory, economically we would say, ASML could jack up the prices by twofold and take what they like.”

  • ASML's revenues have become less cyclical than the semiconductor industry because it is the sole supplier of leading-edge lithography, so customers cannot switch and a customer that cancels an order goes to the back of the queue, though cancellations still occur in downturns.

    “if you cancel your order because things are a bit tough this year, you're to the back of the queue”

  • ASML recognizes it is the bottleneck at the leading edge and invests heavily in capacity so customers do not begin exploring alternative ways of making semiconductors.

    “They don't want to incentivize their customers to start thinking about different ways of making semiconductors.”

  • ASML's competition is not another company but Moore's Law itself, since customers keep buying machines only if they deliver productivity gains at an acceptable price, and the market will switch to alternative paths when lithography cannot keep up.

    “ASML's competition is not really another company, but it's Moore's Law.”

  • ASML's top two customers accounted for nearly 60 percent of revenue last year, and including an undisclosed third customer, the top three likely represent roughly 65 to 70 percent of revenue.

    “So their top two customers were nearly 60% of revenues last year.”

0.38

Moore's Law forecasts are unreliable, so investors should bet on structural chip demand and human ingenuity.

5 pointscentrality 3/5
  • Advances in manufacturing that allow more advanced semiconductors at the same price propagate to end products such as iPhones, NVIDIA processors running more complex algorithms, and AI systems like ChatGPT, making lithography central to computing progress overall.

    “every time they're able to advance the manufacturing process that enables the production of more advanced semiconductors at the same price”

  • ASML's core opportunity is that the world's need for more transistors and chips will keep growing for a very long time unless something dramatic happens, making simple delivery on smaller, sharper, faster lithography a large structural growth story.

    “because the number of transistors we need, the number of chips we need in the world is going to keep growing”

  • Investors should not obsess over perfect entry and exit points in a cyclical industry, because that focus can blind them to ASML's structural opportunity and excellent long-term prospects measured over five, ten, or twenty years.

    “that obsession on the perfect entry and exit point can blind you to the structural opportunity”

  • Even ASML's top technologist, Martin van den Brink, has repeatedly underestimated how long Moore's Law would continue, first predicting at least 15 more years in 1999 and then about another 15 years 15 years later, showing that long-range forecasting in this industry is nearly impossible.

    “Martin van den Brink said to him, I think, at least 15 years.”

  • Investors should not underestimate human ingenuity, since ASML's history shows that the sustained innovation that keeps Moore's Law alive is consistently underappreciated, and even the most expert insiders cannot see five years ahead.

    “not to underestimate the power of human ingenuity”

0.36

ASML's margins, free cash flow, and recurring new-machine revenue show a highly profitable business.

4 pointscentrality 3/5
  • ASML converts nearly all net income into free cash flow, helped by customer down payments, and uses it for dividends, share buybacks, and R&D spending of about 15 to 16 percent of revenue, while still investing heavily in capital expenditure.

    “So pretty much net income converts to free cash flow generation.”

  • ASML's gross margins are now around 50 percent, up from the mid-40s last decade and roughly 30 percent in the 2000s, with operating margin around 30 percent last year, and likely to improve further as next-generation EUV ramps.

    “So their gross margins are now around 50%.”

  • New machine sales account for about 75 percent of ASML's revenue, while the remaining balance comes from service and field options, meaning maintenance and upgrades.

    “their new machine sales account for about 75% of revenues.”

  • In 2022 ASML generated about EUR 21 billion in revenue and about EUR 6.5 billion in operating profit, with a market capitalization just over EUR 250 billion at the time.

    “In 2022, they generated revenues of about EUR 21 billion and operating profits of about EUR 6.5 billion.”

0.27

Lithography is extremely hard to execute, but modular design and incremental milestones make progress manageable.

6 pointscentrality 2/5
  • ASML can leverage computational holistic lithography and e-beam metrology, which fires electrons at chips to find defects at roughly single-nanometer resolution, feeding data back to adjust lithography machines and improve yields until Moore's Law runs out.

    “you need to fire electrons at the surface of a semiconductor to identify where the defects are arising.”

  • Unlike drug approvals, technology milestones at ASML are typically short technical updates, such as power achievements in the light source or mirror development progress for high-NA, rather than binary announcements, because most improvements are incremental.

    “they would report to the market on terms of test achievements.”

  • ASML's modular machine design lets it develop and upgrade individual modules in parallel, so a failed upgrade in one module does not take down the whole machine, which makes continuous improvement highly derisked even though the underlying physics is extremely risky.

    “they just break it down into components and they make each of those in parallel and then put them together.”

  • EUV machines are roughly the size of a double-decker bus and need three jumbo jets to transport, and the next generation will be bigger because making patterns smaller requires larger lenses.

    “Because perversely, as you make the pattern smaller, you have to make the lenses bigger and that means the machine has to get bigger.”

  • Using deep ultraviolet light at 193 nanometers to make 5-nanometer chip features is like writing a signature with a snow shovel: possible but error-prone.

    “So it's a bit like trying to write your signature using a snow shovel or something.”

  • Although the concept of lithography is simple, executing it is extremely difficult because the features are measured in handfuls of atoms rather than millimeters.

    “they're measured in handfuls of atoms rather than in millimeters or centimeters.”

0.22

ASML faces near-term geopolitical and order-cycle risks while expanding its manufacturing capacity.

4 pointscentrality 2/5
  • Several semiconductor companies recently signaled that conditions were tougher than expected, with TSMC cutting revenue guidance, and Tom expects machine order cancellations and pushouts to continue in future cycles.

    “TSMC recently cut their revenue guidance.”

  • Geopolitics is a near-term risk for ASML because last year nearly 40 percent of sales went to Taiwan, nearly 30 percent to South Korea, and around 50 percent to China, a customer base that could become difficult amid tensions.

    “last year, they made nearly 40% of sales were to Taiwan, nearly 30% of sales were to South Korea, and give or take, 50% of sales were into China.”

  • ASML is primarily an architect and integrator: about 80 percent of cost of goods sold is components and materials bought from outside suppliers, and about 20 percent is on-site labor, so capacity expansion means expanding assembly capacity.

    “About 80% of the cost of goods sold is in components and materials that they're buying from outside the business and the remaining 20% is coming from labor actually on site.”

  • ASML's capacity plan targets 90 EUV machines per year, compared with about 40 sold in each of the last two years, plus about 600 deep ultraviolet machines and roughly 20 high-NA EUV machines by around 2027-2028.

    “They are investing so that they can produce 90 EUV machines per year.”