Digital Sovereignty

ASML Fell 4.6 Percent on Five Machines

China has started building its own immersion DUV lithography tools. Five this year, about twenty in 2027. What that changes for anyone buying 28nm parts.

Digital SovereigntyBy Servola Tech Desk2026-07-274 min read

AI-assisted, edited by humans. | Editor-in-Chief: Leon Soliman | Editorial standards

Key takeaways

China has begun producing its own immersion DUV lithography tools and will deliver roughly five this year and about twenty in 2027 to SMIC, Hua Hong and ChangXin Memory, which is not enough to displace ASML but is enough to end the assumption that a Chinese mature-node fab can be switched off by withholding service.
If you buy 28nm or older parts, stop pricing a security-of-supply premium into your 2027 and 2028 contracts on the basis that Chinese mature-node capacity can be politically constrained, because that constraint now has a visible expiry date.
Ask your component distributor which of your part numbers are fabricated in China and at what node, and get the list this quarter, because the exposure that matters is at 28nm and above, not at the leading edge everyone writes about.
Treat the 4.6 percent move in ASML shares as information about consensus rather than about ASML: a five-unit production run repriced Europe's most valuable technology company, which tells you how thinly the belief in a permanent tool monopoly was actually held.

Five tools, three customers, one node that matters

The Information reported on 27 July that China has begun manufacturing domestically developed immersion deep ultraviolet lithography machines, the tools that print circuit patterns onto silicon. Reuters carried the report the same day. The volumes are small and deliberately stated as such: roughly five machines during 2026 and about twenty projected for 2027. The named recipients are Semiconductor Manufacturing International Corporation, Hua Hong Semiconductor and ChangXin Memory Technologies, the memory maker that listed in Shanghai on the same day and closed its debut up around 470 percent.

The engineering caveat is real and it is in the reporting: the systems still lag on performance and reliability and need further testing before genuine mass production. ASML's advantages are intact. None of this touches extreme ultraviolet, the generation above, which remains closed to Chinese buyers under US-led export controls and which no one else in the world builds. The story is not that China caught up. It is that China now builds a working version of the tool one generation below the frontier.

This did not appear from nowhere. SMIC began testing a domestic immersion DUV tool built by Shanghai Yuliangsheng Technology in September 2025, and SMEE has reportedly sold around ten units of its SSA800 series aimed at 28 nanometre immersion work. The 27 July report marks the transition from testing single units to producing them on a schedule.

Why five machines repriced a European champion

ASML shares fell 4.6 percent on the report. Set against a company that is by market value Europe's most important technology business, a five-unit production run is a rounding error in revenue terms. The market did not reprice ASML's order book. It repriced a belief.

The belief being repriced is about servicing, not selling. For three years European and American policy has rested on a quiet assumption: export controls on lithography tools bite twice, once when a fab cannot buy a machine and again, more painfully, when the machines it already owns cannot be serviced, calibrated or spare-parted. An installed base you cannot maintain degrades. That second lever is what made the controls feel decisive, and it is the lever a domestic alternative erodes first, because a fab that can buy a replacement locally can absorb the loss of a service contract in a way it could never absorb the loss of the tool itself. Washington is now weighing tighter restrictions on foreign tool exports and servicing inside China, which is a response to exactly this.

The counter-case deserves stating plainly: twenty machines in 2027 does not build a semiconductor industry, immersion DUV at 28 nanometres is a technology ASML shipped commercially many years ago, and reliability problems in lithography are not minor engineering details but the difference between a tool that runs a fab and a tool that sits in a bay. Every one of those points is correct. They constrain the timeline. They do not restore the assumption that the tool supply is a permanent chokepoint.

What a European industrial buyer should do about it

Start with the parts list, because the exposure is not where the coverage is. Leading-edge logic gets the attention, but the silicon inside industrial controllers, vehicle subsystems, smart meters, appliances and building systems is overwhelmingly mature node, and mature node is precisely the segment this affects. Ask your distributor which of your part numbers are fabricated in China, at what node, and with what qualified second source. Set a deadline of this quarter and put the answer in your bill of materials rather than in an email thread.

Then reprice the premium, not the risk. Many European manufacturers have spent two years paying more for non-Chinese mature-node parts, or holding larger buffers of them, on the reasonable theory that Chinese supply could be interrupted by policy. That theory is weakening on a schedule you can now see. The practical consequence is that the premium should compress over the next several contract cycles, so a three-year supply agreement signed today at security-of-supply pricing may be locking in a cost that the market removes before the term ends.

Finally, be honest with your board about which risk you actually hold. The risk in mature-node semiconductors was never that the parts vanish. It is concentration: a small number of fabs, in a small number of jurisdictions, making components that sit in products with fifteen-year service lives. More capacity in China does not reduce that concentration, it relocates part of it. Qualifying a second source in a different jurisdiction is still the only action that changes your position, and it is worth budgeting for on its own merits rather than in reaction to this week's headline.

Frequently asked questions

Does this mean ASML's business is in trouble?

No, and nothing in the reporting supports that reading. The Chinese systems lag on performance and reliability, the volumes are tiny against ASML's shipments, and extreme ultraviolet lithography, where ASML has no competitor anywhere in the world, is untouched. What changes is a policy assumption about permanent leverage rather than a commercial forecast for one company.

We buy finished modules, not chips. Does this reach us?

Yes, indirectly, and usually invisibly. Module vendors rarely disclose fab origin, so the practical step is to ask them in writing which nodes and which fabs sit behind the components in the modules you buy, and whether a qualified alternative exists. If a supplier cannot answer that question about its own product, that is itself a finding worth recording.

Should we delay contracts to wait for prices to fall?

No. The direction is more predictable than the timing, and mature-node pricing is being pushed in several directions at once by memory demand and by AI-driven substrate and packaging constraints. The better response is structural rather than tactical: shorten the term on new security-of-supply pricing, add a benchmarking clause, and avoid signing multi-year premiums that assume a constraint which is now dated.

Five machines will not build a competitor to ASML, and anyone telling you otherwise is selling something. What five machines do is convert a chokepoint into a countdown, and a countdown can be planned against. Get the fab origin of your mature-node parts on paper this quarter, then price your next supply agreement for the world that countdown describes rather than the one the controls were written for.

ASMLDUV lithographySMICChangXin Memoryexport controlssemiconductor equipmentmature nodes

China Trained a Frontier AI With No US Chips

Meituan trained a 1.6 trillion parameter model end to end on 50,000 domestic Chinese chips. The export-control moat just gained a public counterexample, and Europe's excuse got weaker.

3 min read

Decisions

Chip Stocks Fell 20%. Your Compute Bill Did Not

On 17 July 2026 the semiconductor index closed more than 20 percent below its June peak, a paper bear market. Demand, capex plans and compute prices did not move. What owners should and should not change.

4 min read

Law & Regulation

The EA Buyout Now Hinges on Brussels

The largest take-private in history has cleared EA shareholders and US antitrust. The last real gate is the EU - and not classic merger review, but the Foreign Subsidies Regulation, facing its first blockbuster test.