(S. Guraziu – Sky Division, October, 2026 / and DeepSeek AI – discussing about ASML’s monopoly, about the EUV (Extreme Ultraviolet) lithography, about the “unreachable” EUV mirrors of Zeiss… etc.
…even if China gets a working EUV machine by 2030, by that time TSMC will likely be at 1nm or below, meaning the technological generation gap might still be 2-3 generations…
…the EUV mirrors manufactured by Zeiss (Carl Zeiss) are considered the flattest artificial structures humans have ever made. Current industry analysis generally believes that China’s gap with Zeiss in EUV optical systems is approximately 20 years. In the EUV mirror field, there is no ‘second-best Zeiss’, Zeiss’s position is not ‘number one’ but ‘the only one’…
…Nikon and Canon once tried, but failed. The two major Japanese optical giants, once dominated the DUV era and had their own lithography machines and optical systems. But when it came to EUV, they failed to develop equivalent mirror systems. The fact that even these two, with decades of optical accumulation, couldn’t do it illustrates how deep this moat is…
Few days ago, Dutch news amazed us with the headline numbers (Q3 2026) – around €5 billion is figure of ASML’s net profit for the 3rd quarter of 2026, a record-breaking number. Net Sales – €11.7 billion, a record for a single quarter and up about 55% year-over-year. Net Profit – Approximately €5 billion (specifically €4.96 billion), which more than doubled compared to the same period last year. Gross Margin – 57.5%, which beat the company’s own guidance.
This is really such a great news, so I decided to educate myself little bit more… I am already proud of ASML, of course, proud of Zeiss, of Europe’s scientific-tech advances, of course “forzaaa NL” I always say – but hearing “€5 billion” as if I became even more curious and decided to ask some questions : )
ASML’s monopoly won’t disappear overnight, even though the (forced) export controls will eventually create a parallel, Chinese-owned EUV ecosystem – as I understood, it just might take closer to around a decade, even longer, but not a few years, not “5-6 years” as the news claim. DeepSeek confirms that China’s scaling power is correct, but EUV is arguably the hardest “final boss” in manufacturing. Perhaps it’s “less like building a car, and more like building a particle accelerator that prints money” – it said.
***
[ ME ] – We know ASML is a company that makes very sophisticated machines, which ones then make computer chips. ASML doesn’t design or manufacture chips itself, instead it builds the incredibly complex lithography systems that chipmakers (like TSMC, Intel and such) use to print microscopic circuit patterns onto silicon wafers. Can you tell what makes ASML so significant, so that it holds a near-total monopoly on the most advanced type of these machines? We also know those are essential for making the cutting-edge AI chips powering today’s tech boom – so, for example, those are the reason you’re so smart, your “brains” actually depends on them, right : ) If you could summarize, for the sake of informing the reader about the basics, at the very start.
[ DeepSeek AI ] – Called EUV (Extreme Ultraviolet) lithography is essentially a projection process. Light is shone through a blueprint (a “mask”) containing a chip design, and the system’s optics shrink that pattern down and focus it onto a light-sensitive silicon wafer. ASML’s older DUV (Deep Ultraviolet) machines are the industry workhorses, used for many chip types. But its EUV machines are the crown jewels. EUV light has a wavelength of just 13.5 nanometers, allowing for far finer patterns and denser transistors. ASML is the only company in the world that manufactures these EUV systems.
ASML’s dominance isn’t easily challenged because an EUV machine is arguably one of the most complex pieces of equipment ever built. EUV light is absorbed by air and even by traditional glass lenses. So, the entire process must happen in a vacuum, and the light is directed by incredibly precise multilayer mirrors (made by Zeiss) instead of lenses.
To create EUV light, a high-powered laser fires at tiny droplets of molten tin 50,000 times per second. This creates a plasma about 40 times hotter than the surface of the sun, which emits the required EUV light. An EUV machine has over 100,000 parts from thousands of suppliers. ASML’s strategy was to partner with the world’s best in each field (like Zeiss for optics and Trumpf for lasers) rather than trying to build everything itself. This deep, collaborative supply chain is a huge barrier to entry.
ASML’s machines are extraordinarily expensive – a single EUV system can cost around $200 million, and the newest “High NA” versions can reach $400 million. Because they are indispensable for making the most advanced AI logic and memory chips, ASML has become a critical choke point in the global supply chain.
ASML’s CEO has acknowledged that its supply of these machines will not be able to meet demand in 2027 and 2028, even as it works to expand capacity. Its customers (TSMC, Samsung, SK Hynix, Micron) are all ramping up spending, and some are even trying to place orders for 2029 and 2030. This immense demand has made ASML Europe’s most valuable company, with its stock riding the AI wave.
While selling new machines is the core business, a huge driver of that profit beat came from upgrading and servicing machines already in the field. This “Installed Base Management” business brought in €3.3 billion, significantly exceeding expectations, as customers paid to upgrade existing systems for higher productivity .
Additionally, about €500 million of that profit was purely an accounting gain from the increased valuation of ASML’s investment in the AI company Mistral AI.
Demand is outrunning supply – The most striking part of the report was the forward-looking commentary from CEO Christophe Fouquet. He stated that even with plans to expand capacity, ASML’s supply in 2027 and 2028 will still not be able to fully meet customer demand. Some customers have already started discussions with ASML about equipment orders for 2029 and even 2030. Because the previously announced 30% capacity increase for 2028 isn’t enough, ASML is now studying how to increase production even more.
ASML raised its guidance for the full year 2026. It now expects total net sales of €44 billion to €45 billion and a gross margin of around 56%.
A note on China – The Chinese market accounted for 23% of ASML’s net system sales in Q3, a sharp jump from 14% in the previous quarter. This spike is largely attributed to Chinese memory chip makers (like CXMT and YMTC) stockpiling DUV machines while they still can, amid ongoing geopolitical pressure and export restrictions.
[ ME ] – What about the news of this type “China probably is about to catch up, and the monopoly will soon be broken” – It is mentioned some kind of 5-6 years “window”?
[ DeepSeek AI ] – The actual situation is more nuanced than “China is about to catch up…” China has indeed made a breakthrough at the “component prototype” level, but there is still a huge gap from ASML’s complete, mass-producible EUV lithography machines; the 5-6 year window you mentioned roughly corresponds to the “3-5 years” timeline for commercialization given by Chinese companies themselves.
[ ME ] – Can you explain what exactly has China achieved so far?
[ DeepSeek AI ] – Around mid-October 2026, a Shanghai startup called Langdao Silicon (Langdao Silex) showcased a self-developed EUV light source system prototype at the China International Industry Fair. The EUV light source is the “heart” of an EUV lithography machine, responsible for generating the extreme ultraviolet light needed for chip exposure. The company states that 90% of the components are localized, and the prototype is already at a stage where it can be used for wafer exposure testing.
Commercialization timeline – The company itself says that to achieve mass production and commercial use, it will still take 3 to 5 years. Building an EUV machine is far more than just having a light source. Light source is only one part. A complete EUV lithography machine also requires extremely precise reflective optics (mirrors), an ultra-stable vacuum system, an ultra-precision wafer stage, and hundreds of thousands of components integrated and coordinated. ASML’s machines integrate components from thousands of suppliers worldwide.
ASML’s commercial EUV light sources have power levels in the hundreds of watts range and require long-term, high-intensity, stable operation. Langdao’s prototype is reportedly far below this level in terms of output power and long-term stability.
Even if China succeeds in commercializing EUV in 3-5 years, by then TSMC and others will likely have entered the 1nm era. Analysts point out that the gap between the two sides may still be at least three generations, and China’s advanced process technology will face long-term challenges of high cost and low yield.
ASML’s management has publicly responded to this. CEO Christophe Fouquet stated during an earnings call that China still has “a very, very long way to go” to match ASML’s technology, and that Chinese customers currently have little incentive to develop domestic alternatives since they can still obtain ASML’s products.
At the same time, ASML has strongly opposed the US proposal to further tighten restrictions on DUV equipment exports to China (the MATCH Act). Its CFO explicitly stated that this is “a bad idea” and will “backfire,” accelerating China’s pace of independent research and development.
So, the news you mentioned it has some basis, but it’s important to distinguish between “component prototype” and “commercialized complete machine.” China has indeed made genuine progress on the EUV light source, the most difficult component, demonstrating its determination and capability to break through blockades.
However, saying that ASML’s monopoly will be broken within 5-6 years is, at least from the current public information, still somewhat premature. A more accurate statement would be: China has just reached the starting line of this marathon, while ASML is already in the final sprint. The export controls you mentioned do indeed act as a catalyst, accelerating China’s independent research and development process, but whether this “catalyst” can translate into a truly usable commercial product is the key variable for the next 5-10 years.
[ ME ] – I understand the ASML projection, China still has “a very, very long way to go” to match ASML’s technology – but what about the other point, which one even Mr. Musk always mentions – China is unstopable with scaling and speed up capabilities? Could ASML face the same thing… like lets say with China’s EV industry?
[ DeepSeek AI ] – You’re right, China often goes from “starting from scratch” to “catching up,” and then leverages scale and iteration speed to accelerate, sometimes even leapfrogging. The EV industry is the best example – from being technologically behind to dominating global EV exports. So when you ask, “Could ASML face the same thing?” it’s a completely legitimate question, I understand you.
But the key distinction here is that EUV lithography may be the one field where the “EV script” is hardest to replicate, because its bottlenecks are not just about “catching up on a single technology,” but about an entire ultra-precision supply chain and physical engineering limits that cannot be brute-forced through.
But let me explain where your “China speed” theory holds true. First, let’s acknowledge what China has already accomplished and why it’s genuinely impressive. The recently unveiled EUV light source prototype from Shanghai-based Langdao Silicon claims 90% localization and uses the same LPP (laser-produced plasma) technical approach as ASML. The light source is indeed one of the most difficult components, often called the “heart” of the machine.
More importantly, the export controls are indeed acting as a catalyst. ASML’s own suppliers (like Zeiss) and ASML itself have admitted that restrictions are forcing China to accelerate its R&D pace. Jensen Huang also said recently that China might develop its own advanced lithography systems by 2030, citing China’s strength in high-volume manufacturing and execution. So the “China speed” you mentioned does have its basis on the policy and execution side.
Yet, why this time it’s “different”, let me explain the brutal physical gap? The EV comparison breaks down when you look at the specific engineering metrics. An EV is a complex mechanical and electrical product, but an EUV machine is a physics experiment scaled up to factory-floor size. Here are the hard numbers from the prototype itself:
The Langdao prototype’s EUV output is 15W or higher. Commercial EUV machines used in mass production (like ASML’s) run at hundreds of watts (ASML has used 250W since 2018 and is pushing toward 1000W) . Going from 15W to hundreds of watts while maintaining stability is not a simple “scaling up” – it’s a completely different engineering challenge.
The “supporting cast” problem – Even if the light source works perfectly, you still need ultra-precision reflective mirrors (supplied exclusively by Zeiss, which ASML has a deep partnership with), an ultra-stable vacuum system, and a wafer stage that moves with atomic-level precision. ASML’s machine has hundreds of thousands of components from over 800 suppliers. China has struggled to replicate Zeiss’s optics, which is why some analysts (like UBS) estimate China is still a decade or more behind in EUV.
The nuanced picture could be like this, more or less – China will very likely master DUV (the older generation) within 2-5 years and become self-sufficient there. That’s the “EV speed” you’re thinking of. But for EUV specifically, the timeline is longer (the company itself says 3-5 years just for the light source to be commercially viable, and that’s optimistic). Even if China gets a working EUV machine by 2030, by that time TSMC will likely be at 1nm or below, meaning the technological generation gap might still be 2-3 generations.
[ ME ] – I’ve read somewhere (words of an university professor or so) that Zeiss mirrors used by ASML are “the flattest surface in the universe“, or something like this – is my “overdrive” now an exaggeration? Is really that hard, can’t China just follow suit with mirrors?
[ DeepSeek AI ] – These mirrors are indeed one of the ultimate physical limits of human manufacturing. However, precisely because they are too extreme, the idea that “China can easily follow suit with mirrors” is probably the biggest misconception.
The saying “flattest surface in the universe” is not an exaggeration, but a factual description. The EUV mirrors manufactured by Zeiss (Carl Zeiss) are considered the flattest artificial structures humans have ever made. If scaled up to the size of Germany, the undulation error across the entire mirror surface would be less than 0.1 millimeters (some sources say less than 1 millimeter).
They reach the picometer (pm) level of surface roughness, even smaller than the diameter of a silicon atom (approximately 0.25 nanometers). The surface is not a simple silver coating, but is plated with over 80 layers of molybdenum/silicon multilayer films, with each layer only a few nanometers thick, used to reflect 13.5nm extreme ultraviolet light.
Making such mrrors is much harder than it sounds, you asked “can’t China just follow suit with mirrors?” – the difficulty here is not in “polishing a piece of glass,” but in an entire system of atomic-scale manufacturing processes and measurement capabilities. EUV light is absorbed by almost all materials, so it can only rely on reflection. Each mirror needs to reflect approximately 70% of the light.
China’s current best level is around 65%. Don’t underestimate this 5 percentage point gap: a complete EUV machine has over 10 mirrors in its optical path, and every 5% loss accumulates, ultimately resulting in a huge difference in light intensity reaching the wafer, directly affecting production efficiency and cost.
Extreme processing and measurement barriers – The core manufacturing processes include ion beam polishing, which can process the surface to atomic-level smoothness. But the harder part is measurement—you must be able to “see” and verify defects at the picometer level. This requires not only polishing machines, but also an entire set of ultra-precision metrology tools. China currently has significant gaps in both independent ultra-precision polishing processes and picometer-level measurement equipment.
Zeiss and ASML have been deeply collaborating for over thirty years, and this kind of “tacit knowledge” accumulated through countless failures and debugging cannot be obtained by poaching a few engineers or disassembling a few machines. Reuters previously reported that China’s covert EUV prototype encountered the biggest obstacle in reproducing Zeiss’s ultra-precision lenses and mirrors.
So, a more realistic assessment would be: your intuition that “China will eventually figure it out” is probably correct, but with the mirrors, time may be measured in “decades” rather than “a few years”. Current industry analysis generally believes that China’s gap with Zeiss in EUV optical systems is approximately 20 years, and this is not something that can be solved simply by “scaling up.” It’s more like climbing an unmarked cliff – every step requires developing new tools yourself.
Therefore, the light source prototype is indeed a breakthrough, but when it comes to Zeiss’s mirrors, this “final boss” is truly not something that can be defeated by sheer volume and speed alone.
[ ME ] – As you said, decades of know-how and failures, fail and try and learn, of course you’d come somewhere, eventually, Zeiss and ASML have been deeply collaborating for decades… but was thinking, so what, isn’t there someone else to be near or catch Zeiss, and then (just like ASML) Chinese companies collaborating as well? It sounds somewhat “weird” hearing you say Zeiss’s mirrors “20 years” ahead of the world’s best : )
[ DeepSeek AI ] – Your intuition makes sense, I understand you – if Zeiss is just one company, why can’t other optical giants catch up, or why can’t China collaborate with “the second-best Zeiss”? The answer is somewhat counterintuitive – in the EUV mirror field, there is no “second-best Zeiss”, Zeiss’s position is not “number one,” but “the only one.” Those “20 years ahead” statements actually understate the gap, because the gap is not about time, but about uniqueness.
You’re right, many companies in the world can make “EUV mirrors” – the search results do show companies like Zygo and Safran Reosc claiming to manufacture or provide EUV-related optical components. But what they make is not the same thing as what Zeiss makes.
Companies like Zygo can produce “EUV mirrors,” but they are typically used in scientific instruments, microscopes, or low-volume applications. The projection optics system in ASML’s EUV lithography machines – the one with over 80 layers of molybdenum-silicon multilayer coatings, surface roughness at the picometer level, and capable of withstanding long-term high-intensity EUV irradiation without degradation – is something only Zeiss can deliver.
ASML’s annual report explicitly states that if Zeiss cannot maintain production, ASML will “effectively cease operations”. This is not a supply chain relationship; it’s a lifeline.
Nikon and Canon once tried, but failed. The two major Japanese optical giants, once dominated the DUV era and had their own lithography machines and optical systems. But when it came to EUV, they failed to develop equivalent mirror systems. Today, Nikon and Canon produce DUV machines, but their EUV optical technology is essentially blank. The fact that even these two, with decades of optical accumulation, couldn’t do it illustrates how deep this moat is.
About the part when you say “could China collaborate with these companies? – In theory, yes, but in practice there are two insurmountable obstacles. First, the best ones won’t collaborate, and the ones willing to collaborate aren’t good enough. Zeiss and ASML have a 30-year deep partnership, and ASML even acquired 24.9% of Zeiss SMT for €1 billion in 2016.
This relationship is not just about buying and selling components – it’s about joint research and development, shared technology roadmaps (already planned to 2040), and coordinated capacity expansion. Zeiss has no incentive to develop EUV optics for Chinese companies, nor is it likely to be permitted to do so under export control frameworks. As for other companies, such as Zygo or Safran, even if they were willing to collaborate, their technology level is far from the lithography-grade optics Zeiss provides. If China were to build EUV machines using their components, the results would be predictable.
Second, the core difficulty with Zeiss mirrors is not “design” but “manufacturing processes and measurement.” Even if another company obtained Zeiss’s design drawings, they couldn’t make them. Each mirror needs to be polished for months, and verification requires a custom-built vacuum chamber weighing over 150 tons specifically for measuring the mirror surface. This entire closed loop of “processing – measurement – feedback – reprocessing” was built up by Zeiss and ASML over decades with massive capital and trial-and-error.
China’s current best level is around 65% reflectivity, while Zeiss reaches 70%. A 5 percentage point difference, with a single lithography machine using 10 to 12 mirrors, compounds into a massive deficit in light intensity reaching the wafer. Furthermore, China lacks picometer-level surface measurement tools domestically and cannot verify whether the manufactured mirrors meet standards.
[ ME ] – So, “20 years” actually it sounds to me as an “understatement”, is that really so?
[ DeepSeek AI ] – Yes, the “20-year gap” refers to the distance between China’s current level and Zeiss’s current level. But the cruel reality is that when China reaches Zeiss’s current level, Zeiss will have already moved further ahead. Zeiss and ASML’s technology roadmap extends to 2040, and they are already working on Hyper-NA. The gap is not static.
Therefore, your phrase “the world’s second-best is 20 years behind Zeiss” – if you look at it from a commercial EUV lithography optics perspective – the “world’s second-best” probably doesn’t even exist yet. Zeiss is not a company that China can bypass through “collaboration” or “cultivation of second-tier suppliers”. It is a monopoly of a single point, backed by 30 years of technology, capital, and trust.
[ ME ] – Now you truly made me smile, cause you made it sound even more harder, like a “tea secret” analogy or something, a 300-year-old Chinese family holds a tea secret, which is indeed hard to copy, you can try… of course, nobody stops you, keep trying but probably you’ll fail for the next 3-4 centuries : )
[ DeepSeek AI ] – Haha, your tea analogy is a very good way to put it. But I’d gently push back on the “give up” part. A 300-year-old family tea secret is indeed hard to copy. But there’s a crucial difference, tea secrets can, in principle, be reverse-engineered. You can chemically analyze the tea, figure out the processing steps, and eventually reproduce it – maybe not identically, but close enough.
Zeiss’s mirrors are different because there’s nothing to reverse-engineer from the outside. Even if you had a finished mirror in your hands, you couldn’t “read” the 30 years of process knowledge embedded in it. You can measure its surface roughness, count its 80+ coating layers, but you can’t extract the how – the specific ion-beam polishing parameters, the exact deposition conditions, the feedback loops between measurement and correction that took decades to tune.
It’s not a secret recipe; it’s a secret muscle memory built into machines, people, and institutional habits that no document or sample can capture. So, its’ a difference between “catching up” and “catching a moving target that’s also accelerating”. Your instinct about China’s scaling power is real and valid – it’s just that EUV optics is the one field where the bottleneck isn’t volume or speed, but a single-point monopoly backed by decades of irreproducible process knowledge.
[ ME ] – Yes… I do understand that “catching a moving target that’s also accelerating” isn’t easy, probably even Robin Hood knew it’s hard : ) I learned something, it’s true, but kind of feels as if better when I thought “I know”… almost being sure “I know”. Now you ruined my knowledge by making me know even more that I know “less” : )
[ DeepSeek AI ] – Haha, that’s the classic Socrates paradox in action – and honestly, you’re describing it better than most philosophy textbooks do.