The Machine That Makes No Economic Sense: China, ASML and EUV Independence

China has begun producing domestically developed immersion deep-ultraviolet lithography machines. It is also testing something considerably harder: an extreme-ultraviolet prototype that can generate EUV light but, when Reuters reported on the project in December 2025, had not yet produced a working chip. Neither development puts ASML's technological leadership in immediate danger. They do, however, make a commercially irrational project increasingly rational.

For almost any normal company, recreating ASML would be an absurd business proposition. The Dutch semiconductor-equipment manufacturer has spent decades developing machines of extraordinary complexity. A European Commission Joint Research Centre report describes the EUV lithography supply chain as containing about 100,000 parts from more than 5,000 suppliers, combining precision optics, high-energy lasers, vacuum engineering, nanometre-scale motion control, metrology and software. Its performance also depends on specialised suppliers, most famously Zeiss for the optical system, whose capabilities developed alongside ASML over decades.

A competitor would have to reconstruct much of this ecosystem, spend enormous sums learning how to make it work reliably, and then sell the result against an incumbent with a better product and a formidable head start. Under ordinary commercial conditions, there is little reason to attempt it. China is attempting it anyway because export restrictions have changed the return on reproducing the technology. A machine that would be a commercial failure in an open market can become strategically valuable when access to the superior alternative depends on another government's permission.

Editorial image generated by the author.

From EUV Light to a Production Line

The recent Chinese advances need some perspective. In July 2026, Reuters reported that state-owned Shanghai Aishengna had begun producing domestically developed immersion DUV lithography machines. Initial volumes are tiny by ASML standards, and producing machines is not the same as proving that they can sustain semiconductor manufacturing. Analysts highlighted the gaps that ultimately matter inside a fab: overlay accuracy, throughput, reliability and performance across thousands of wafer runs.

The Chinese EUV effort is further from commercial maturity. Reuters reported that a prototype completed in Shenzhen in early 2025 had become operational enough to generate EUV light. It had not yet produced working chips. The Chinese government's reported target is to achieve that by 2028; people close to the programme regarded 2030 as more realistic.

The distinction is more than semantic. EUV depends on producing 13.5-nanometre light by firing powerful lasers at droplets of tin, directing the resulting light through extraordinary reflective optics, moving wafers at speed while maintaining extreme positional accuracy, and keeping contamination under control in a vacuum. Each layer must align correctly with those already printed. An impressive laboratory result can therefore remain a long way from a machine to which a semiconductor manufacturer would entrust a production line.

ASML's own history demonstrates how large that distance can be. Prototype EUV tools existed years before EUV became attractive for volume manufacturing. The difficult engineering did not end when the machine first worked. Reliability, throughput, source power, contamination control and process integration had to improve until using EUV made economic sense for the fab as well as technical sense for the engineer.

Why Twenty Years Is Not a Twenty-Year Gap

China can attack those problems with resources few ordinary commercial challengers could contemplate. Beijing can fund competing research programmes, subsidise suppliers, absorb losses and provide domestic customers willing to work with immature systems. It can coordinate universities, research institutes, semiconductor manufacturers and state-backed industrial groups around a capability judged strategically important.

Capital can accelerate a great deal, but some industrial knowledge emerges only through repeated use. A component behaves differently after thousands of operating hours. A vibration irrelevant in a prototype becomes destructive when throughput rises. A material that meets its written specification may still cause reliability problems months later. Engineers learn which tolerances really matter, which suppliers can repeatedly meet them and which apparently minor imperfections propagate through an entire production process.

That is why describing China as simply "ten years behind" or "four generations behind" can mislead. Technological gaps are not measured cleanly in calendar years. Ten years of accumulated industrial experience cannot necessarily be purchased in two by multiplying the research budget.

But the opposite conclusion is equally suspect. ASML and its partners had to discover which technological routes would work. China begins with the knowledge that a successful architecture exists. It can study patents and scientific literature, analyse imported equipment, recruit experienced engineers and reverse-engineer components. Reuters' investigation found former ASML engineers involved in the Chinese programme and reported that older equipment and second-hand components had contributed to the prototype.

A follower can avoid branches of the technological tree that the pioneer already proved barren. It does not have to rediscover why EUV requires a particular type of source or why reflective optics are necessary before trying to manufacture them. That advantage does not remove the problem of reproducing Zeiss-class optics, achieving industrial uptime or building a supplier base capable of meeting extreme tolerances year after year. It does mean that ASML's development history should not be treated as a stopwatch that China must rerun from zero.

Significant catch-up is therefore plausible without being easy to date. The harder question is how much accumulated industrial knowledge can be compressed and how much China will still have to acquire one failed production run at a time.

The Target Keeps Moving

Technological catch-up is often described as though the leader waits politely at the finish line. ASML does not. While China works towards capabilities comparable with earlier generations of Western lithography, ASML continues improving source power, throughput, overlay and the wider manufacturing process around its machines.

ASML demonstrated a 1,000-watt EUV light source in 2025, a step towards considerably higher future throughput. The frontier has also moved beyond conventional EUV. In July 2026, ASML announced that Intel was using High-NA EUV on selected Intel 18A layers in high-volume manufacturing. By 8 September, ASML and Intel reported that High-NA EUV had processed more than one million wafers across early tool certification and testing, research and development, and volume production, with overlay, throughput and availability meeting Intel Foundry's expectations. China is therefore chasing an incumbent that continues to move the definition of leading-edge lithography forward.

There are at least two different objectives hidden inside the phrase "catching ASML." One is to reproduce a sufficiently capable EUV system. The other is to reach the technological frontier while ASML continues advancing it. The first may prove considerably easier than the second, and Beijing does not necessarily need the second to achieve its strategic objective.

A domestic lithography ecosystem permanently several years behind ASML could still remove much of China's present vulnerability. Its machines might expose fewer wafers per hour, require more maintenance and make leading-edge fabrication more expensive. Chinese manufacturers could remain at a productivity disadvantage while retaining the ability to make sufficiently advanced chips without relying on equipment subject to foreign export licences. Technological independence does not require technological leadership.

When an Inferior Machine Becomes Valuable

Suppose a future Chinese EUV system costs substantially more to operate than its ASML equivalent. It processes fewer wafers, suffers more downtime and requires additional engineering support. In an unrestricted market, a semiconductor manufacturer would have little reason to buy it. Billions might have been spent producing a worse machine.

That comparison changes once the better machine cannot be purchased. China has never received ASML's EUV systems, and restrictions also cover some advanced DUV equipment. Chinese manufacturers are not completely without alternatives. DUV can be pushed further through multi-patterning, splitting complex patterns into several exposures and additional processing steps. The technique predates EUV and can extend older lithography surprisingly far, but at the price of additional complexity, equipment time and process steps.

The workaround is not merely theoretical. On 20 September 2026, Chinese memory manufacturer CXMT said its fifth-generation DRAM platform had entered mass production using quadruple patterning, reducing the spacing of key memory-cell features to 11.95 nanometres. That does not amount to replacing EUV for leading-edge logic, but it illustrates how additional patterning can extract more capability from available manufacturing technology.

The strategic choice is therefore not literally Chinese EUV or no advanced chips. It is an inferior domestic system against increasingly elaborate workarounds, continued reliance on whatever foreign equipment remains available, and the risk that future restrictions make those routes more difficult.

Under those conditions, the economics change. A machine that cannot beat ASML in a normal procurement competition may still be worth building because much of its value sits outside the machine manufacturer's accounts. Advanced semiconductor manufacturing supports artificial intelligence, telecommunications, weapons systems, scientific computing and industrial automation. Domestic lithography reduces the possibility that access to those capabilities can be constrained by foreign policy.

China's programme consequently looks less like an ordinary technology investment than strategic infrastructure. A government may rationally pay far more for domestic capability than a private buyer would because it is purchasing something the market price of the machine does not contain: the option to retain access when international supply fails.

Export Controls Change the Return on Duplication

This creates an awkward problem for Western technology policy. ASML's position is protected by extraordinary technological barriers, but also by something more mundane: reproducing its ecosystem has historically made little economic sense. Why spend tens of billions building an inferior lithography industry when the world's best supplier will sell you a better machine?

Global specialisation reinforced ASML's position because buying from the specialist was cheaper than recreating what the specialist knew how to do. Export controls change that calculation. They make catch-up harder by denying China access to leading equipment, components and expertise. That delay can be strategically valuable. In technologies advancing as rapidly as artificial intelligence and advanced semiconductors, even several years of constrained access may have significant consequences.

The same restrictions, however, increase the value of overcoming the barrier. Once continued access to the superior foreign machine can no longer be assumed, the economic absurdity of recreating it begins to disappear. China also gains something almost no conventional competitor to ASML would enjoy: a large domestic market with strategic reasons to tolerate immature equipment.

An early Chinese immersion lithography system does not need to persuade TSMC or Samsung that it is better than ASML. Domestic fabs can use it because establishing an independent supply chain has value of its own. Operating those systems then generates the failures, measurements and production experience from which the next generation can improve. At the same time, Chinese chipmakers continue purchasing less advanced foreign DUV equipment where regulations permit, so the transition towards self-sufficiency is likely to be gradual rather than a clean break.

Export controls therefore do something more complicated than either of the simplest political narratives allows. They do not merely "fail" because China develops substitutes, nor can they permanently freeze Chinese technology at its present level. They impose a genuine technological cost while simultaneously increasing the economic value of learning to escape that cost.

ASML may remain far ahead throughout that process. It could remain the world's preferred lithography supplier indefinitely. China does not need ASML to lose for its own programme to succeed.

The Price of Strategic Independence

The argument extends beyond lithography. For several decades, international production increasingly concentrated around companies and countries capable of performing particular tasks exceptionally well. That specialisation lowered costs and accelerated technological progress, but it also created dependencies that were easy to overlook while access was taken for granted.

Strategic competition changes the calculation. Efficiency still matters, but it can no longer be the only question. Governments also have to ask what happens if the most efficient supplier becomes unavailable and how long rebuilding an alternative would take.

For Europe, the same calculation sits inside the broader question of whether it can still build and scale strategic industrial capacity. European policymakers frequently want autonomy in semiconductors, artificial intelligence, cloud infrastructure, defence, energy and critical materials while also expecting the resulting industries to be privately profitable and competitive with established foreign suppliers. Sometimes those objectives align. In harder cases they do not.

Capacity that looks redundant under ordinary cost accounting may partly represent an insurance premium against losing access altogether. That does not turn every subsidy into sensible industrial policy. "Strategic autonomy" can easily become a convenient label for preserving politically favoured industries whose disappearance would create little genuine vulnerability. The difficult task is deciding which dependencies are dangerous enough to justify paying for duplication and which merely provide an excuse for permanent protection.

Advanced semiconductor manufacturing sits relatively close to the easier end of that judgment for China. Its role in economic and military power is large enough that allowing foreign governments an effective veto over critical production equipment creates an obvious vulnerability. Beijing has shown that it is prepared to spend heavily to reduce it.

The eventual Chinese EUV machine may never be as good as ASML's. It may cost more, operate less reliably and remain generations behind the frontier. Chinese fabs may continue paying a substantial productivity penalty for using domestic equipment. None of those outcomes necessarily means that the programme failed.

If China can eventually manufacture sufficiently advanced chips through a supply chain that foreign governments cannot switch off, the strategic objective has largely been achieved. The usual language of technological competition obscures this because it assumes China must build a better ASML. More plausibly, China is trying to build a world in which it no longer needs ASML.

A technological monopoly is protected not only by the difficulty of reproducing the technology, but by the irrationality of doing so. Once access becomes uncertain, that irrationality begins to disappear. The greatest threat to a technological monopoly may therefore not be a company capable of building a better product, but a state willing to build a worse one indefinitely because dependence has become more expensive than inefficiency.