China’s EUV Gap Is Still About 15 Years, Even as Domestic Immersion DUV Ramps Up

China’s push for semiconductor self-sufficiency has reached an important new stage: domestic companies are reportedly moving immersion DUV lithography tools into low-volume production, while CXMT has begun small-scale HBM3E output for Chinese AI-chip developers. Those are consequential milestones. They should not, however, be confused with China having closed the gap in leading-edge EUV lithography.

The clearest distinction is between immersion DUV, which can support increasingly capable domestic manufacturing with technical trade-offs, and EUV, the much harder platform at the center of the world’s most advanced logic-chip production.

The EUV verdict: progress is real, but the gap remains large

Frank Rohmund, head of Zeiss’s semiconductor business, said it is a “reasonable assumption” that China could need about 15 years to develop its own EUV lithography equipment, while cautioning that the pace of progress is difficult to quantify. UBS reached a similarly cautious conclusion, characterizing China’s lithography capability as roughly comparable with ASML’s 2004 level and indicating that domestic EUV replacement is unlikely within the next decade.

That is not a forecast that China will stand still for 15 years. It is an assessment of how difficult it is to turn a working concept into a dependable, repeatable manufacturing platform with the required optics, light source, wafer handling, software, metrology, service network, yields, and production capacity.

Why ASML and Zeiss are such a difficult combination to replicate

ASML is the only commercial manufacturer of EUV lithography systems. Zeiss is its exclusive supplier of the optical equipment used in those systems.

EUV is essential to producing the most advanced chips, including processors used in Nvidia AI accelerators. Zeiss describes its EUV optical systems as ultra-precise mirror-and-mechatronic assemblies that form a key component of EUV machines.

This is why the challenge is larger than designing a single scanner. A competitive EUV platform depends on a specialized industrial ecosystem that can manufacture and integrate extraordinarily precise components, then sustain them in high-volume fab operation.

Why EUV matters even when immersion DUV can go further

Immersion DUV remains a strategically valuable technology. It uses a layer of purified water between the lens and wafer to improve resolution; with multiple patterning, it can be extended beyond the feature sizes available in a single exposure. 14

But that route involves more manufacturing steps than using EUV for the relevant layers. The practical issue is not simply whether a chip can be demonstrated using DUV techniques. It is whether a manufacturer can make advanced chips at competitive cost, yield, speed, and scale. That is the advantage EUV offers for leading-edge logic and power-efficient AI hardware.

China’s immersion-DUV milestone is early, low-volume production

Reuters reported that a state-backed Shanghai program had begun producing domestic immersion-DUV machines, with planned output of about five systems in 2026 and roughly 20 in 2027. The first reported customers include SMIC, Hua Hong, and CXMT. 14

Those figures make the significance—and the limitation—clear:

• Significance: China is building a domestic option for a tool class that is central to advanced chipmaking and has long been dominated by foreign suppliers.
• Limitation: five systems in a year is a pilot-scale ramp. Reports say the first units are intended for production-line validation and still require further testing, rather than representing mature, high-volume fab deployment. 10

Reports also require care in assigning credit. The new immersion-DUV effort has been associated with a Shanghai state-backed manufacturer and with Yuliangsheng testing at SMIC, while SMEE is separately reported to mass-produce 90 nm-class DUV systems and to be developing a 28 nm immersion model. Public reporting does not yet establish broad, independently verified high-volume deployment of SMEE’s 28 nm immersion platform. 1411

In short, “mass production” here should be read as the start of limited manufacturing, not parity with the throughput, reliability, overlay performance, installed base, or supplier ecosystem of ASML’s established systems.

Export controls: a constraint that also sharpens the incentive to localize

U.S.-led export restrictions have blocked China’s access to ASML’s EUV systems. They have also increased the strategic importance of domestic equipment programs.

Rohmund’s position is that extending restrictions further to legacy DUV tools could be counterproductive. In his view, additional controls would accelerate innovation in China rather than remove its incentive or ability to pursue alternatives.

That is an argument about incentives, not a claim that restrictions are irrelevant. Controls can restrict access to current best-in-class equipment; the counterargument is that cutting off more mature tools may encourage faster investment in domestic substitutes, particularly in the large DUV market.

CXMT’s HBM3E is a separate—but related—AI supply-chain advance

China’s AI-chip ambitions also depend on high-bandwidth memory, not only logic fabrication. CXMT has reportedly started small-scale HBM3E production, with Chinese AI-chip developers including Alibaba’s T-Head and Cambricon testing the memory. The company is reported to be targeting a larger production ramp in 2027. 1920

This is a meaningful advance because it could create a domestic source of advanced memory for Chinese AI accelerators. Yet the reported status remains small-scale production and customer qualification. Source reporting describes CXMT as about one product generation behind products already being mass-produced by the leading global memory makers, and its capability at scale remains unproven. 20

The bottom line

China is no longer only pursuing lithography self-sufficiency in laboratories and roadmaps. Reported low-volume immersion-DUV output and CXMT’s HBM3E trials suggest that portions of a domestic advanced-chip supply chain are beginning to move toward deployment.

But the EUV gap is fundamentally deeper. ASML’s unique scanner position and Zeiss’s exclusive EUV optics role illustrate why an indigenous EUV machine requires much more than a single technical breakthrough. Based on Rohmund’s and UBS’s assessments, China’s path to a competitive, mass-producible EUV platform remains measured in years—potentially more than a decade—rather than in the current DUV production cycle.