China's EUV Light Source Hits 150 Watts; Standard Is 500

The light source inside China's most advanced extreme ultraviolet lithography prototype produces between 100 and 150 watts of 13.5-nanometer radiation. The industry standard, supplied by ASML subsidiary Cymer for every commercial EUV scanner on the planet, runs at 250 to 600 watts. That gap is not a rounding error. It is the difference between a tool that can pattern wafers at economically viable throughput and one that cannot.

Noah Tan, writing in The Diplomat on 2 July, identified three specific technical chokepoints that determine whether any nation can field a working EUV lithography system: the high-power 13.5-nm light source, mirrors polished to atomic-scale flatness for reflecting that light, and ultra-pure photoresist chemicals capable of converting the EUV pattern into a physical stencil on silicon. China's position on each axis is measurable, and the measurements are not ambiguous.

On mirrors, China has achieved 65 percent reflectivity. Carl Zeiss SMT, ASML's exclusive supplier, ships optics with 70 percent reflectivity for the TWINSCAN NXE series. A five-percentage-point shortfall in mirror efficiency compounds across the roughly one dozen mirrors in an EUV optical train, producing a total throughput penalty that makes volume manufacturing uneconomical. On photoresists, The Diplomat notes that China "cannot produce viable photoresist" for EUV lithography at all, against an industry requirement of sub-parts-per-billion to parts-per-trillion metal impurity levels.

Reuters reported in December 2025 that researchers in Shenzhen had secretly assembled an EUV prototype. The revelation triggered a wave of commentary about how close China might be to one of the last remaining obstacles to indigenous advanced semiconductor manufacturing. The answer, according to The Diplomat's synthesis of industry and policy sources, is that Chinese insiders describe 2030 as a "realistic target" for producing working chips, while skeptics outside the country argue it could take decades to reach commercial viability. Both camps are guessing. The chokepoints are not.

Paul Triolo, technology policy lead at Albright Stonebridge Group, told The Diplomat that industry observers believe an EUV prototype has been "complete" for almost two years. In a bullish scenario, Triolo suggested, Chinese companies could develop a production-ready system within a timeframe that would surprise Western policymakers. The talent pipeline feeding those efforts is real: Lin Nan, who now leads light-source research at the Shanghai Institute of Optics and Fine Mechanics, previously worked on light source technology at ASML. Huawei has repeatedly attempted to recruit precision optics engineers from Zeiss with salary packages designed to overcome non-compete constraints.

Then, on 19 June 2026, the White House introduced a counterpoint that landed somewhere between a warning shot and a mystery. Bloomberg reported that Commerce Secretary Howard Lutnick had, in a series of recent meetings, told senior ASML executives he was concerned that one of the Dutch company's extreme ultraviolet lithography machines may have reached China. ASML's response was immediate: the company "said on Friday it had never shipped an extreme ultraviolet (EUV) lithography machine to China," Reuters confirmed.

Connie Loizos at TechCrunch unpacked the commercial logic behind ASML's denial. An EUV scanner is not a crate of components that vanishes into a grey-market shipping lane. Each TWINSCAN NXE:3800E weighs approximately 180 tonnes, ships in 40 freight containers, and requires a dedicated cleanroom built to class-1 particulate specifications. ASML installs and calibrates every machine on-site with its own engineers, who remain embedded for months. If a system were diverted post-shipment, the service logs, the software telemetry, and the installation team's location records would all have to be falsified simultaneously. ASML would lose its export license and, with it, roughly 30 percent of its annual revenue.

The South China Morning Post framed the episode differently on 24 June. Even if the specific rumour was easy to dismiss, it exposed a structural tightening that Chinese fabs are already living with. "If EUV is out of reach and access to DUV is tightening, there is a limit to how far chipmakers can go with older gear and new approaches," the SCMP noted. The Netherlands, under US pressure, has progressively narrowed the scope of deep ultraviolet (DUV) immersion tools that can be exported to China without a licence. SMIC's most advanced node, publicly acknowledged as its second-generation 7nm process, was built with DUV immersion scanners using multi-patterning. Each additional patterning step reduces throughput, increases defect density, and lowers yield. The economics degrade geometrically.

ASML, meanwhile, has never been stronger. In the first quarter of 2026, the company posted revenue of $10.34 billion with 53 percent gross margins, according to 24/7 Wall St. Its stock had surged 77 percent year-to-date and 142 percent over the trailing twelve months. In early June, ASML became the most valuable company in European history, exceeding $668 billion in market capitalisation and passing the record Novo Nordisk set in mid-2024. Elon Musk called it "arguably the greatest company in Europe."

In an interview with TechCrunch in May, ASML CEO Christophe Fouquet addressed competitive threats directly. "No one is coming for us," Fouquet said, a statement that functions less as bravado than as a description of the capital-goods supply chain that makes EUV possible. ASML spends €4.5 billion annually on R&D. Its newest High-NA EUV platform, the TWINSCAN EXE:5200, costs roughly $400 million per unit, a figure that TSMC executives have described as too expensive even as they remain the only customer with a clear roadmap for using the tool at scale. The company's supply chain reaches backward into a single German optics firm, a single San Diego light-source subsidiary, and a precision-mechatronics ecosystem in the Netherlands that has no duplicate anywhere.

But the equipment story in mid-2026 is not only about lithography. New Electronics reported on 30 June that the global surge in semiconductor fabrication investment is now placing growing strain across the entire tooling supply chain. The Semiconductor Equipment and Materials International consortium projected that worldwide fab equipment spending would exceed $140 billion in 2026. That capital is flowing into logic, DRAM, and NAND capacity simultaneously, driven by AI-training clusters, HBM4 memory stacks, and on-device inference silicon. Every category of tool, from atomic layer deposition chambers to plasma etchers to wafer-bonding systems, faces extended lead times.

The Supply Chain Management Review published an analysis in its July 2026 issue examining what it termed "lead time economics" across the semiconductor supply chain. The core finding: equipment lead times that had normalised to 6-9 months in 2024 have begun stretching again toward 12-18 months for certain categories, particularly vacuum-related subsystems and advanced process-control metrology tools. The bottleneck is not raw materials. It is the specialised engineering labour required to assemble, calibrate, and install tools whose tolerance specifications are measured in angstroms.

The metrology and inspection equipment segment illustrates the squeeze. A Yahoo Finance UK market report published on 30 June described the segment as facing "growing opportunities and challenges amid rising complexity and defect sensitivity in device architectures." As gate-all-around transistors replace FinFETs at the 2nm node and beyond, the number of inspection steps per wafer increases, sometimes by 40 percent or more. KLA Corporation and Applied Materials dominate this segment. Both have reported order backlogs extending into 2027.

Amy Fan, reporting for DIGITIMES from Taipei, detailed how the HBM4 memory transition is generating a parallel wave of equipment demand in South Korea. Samsung Electronics and SK Hynix accelerated investment in advanced DRAM and high-bandwidth memory in the first half of 2026, driving orders across front-end tools, advanced packaging lines, and inspection equipment. The shift toward hybrid bonding for HBM4 stacks, which requires new wafer-bonding tools from suppliers such as EV Group and Tokyo Electron, is creating an additional bottleneck layer distinct from the lithography-centric supply chain that dominates EUV coverage.

Ron Schmelzer, writing in Forbes in May, identified labour as the under-discussed constraint. "The insatiable demand for AI and in turn the demand for compute with huge memory requirements is turning AI's boom into urgent supply chain and labor problems," Schmelzer wrote. A single advanced fab requires roughly 1,500 process engineers and equipment technicians to operate. TSMC alone is building two new fabs in Arizona, one in Japan, and one in Kumamoto, while Samsung expands in Taylor, Texas, and Intel pushes forward with its Ohio site. The global pool of engineers who can bring up a 2nm-capable fab is measured in the low thousands.

What about threats to ASML's lithography monopoly specifically? Pat Gelsinger's stealth startup, xLight, has been working on an alternative EUV architecture since his departure from Intel. 24/7 Wall St described xLight as a "lab-stage startup targeting first prototype validation in 2028." That timeline, even if met, puts a competitive product no sooner than the early 2030s. By then, ASML expects to have shipped multiple generations of High-NA systems and to be qualifying its hyper-NA platform, targeting 0.75 numerical aperture. The startup's existence is a signal that the market wants a second source. It is not yet a supply.

The real bottleneck in the EUV supply chain is not the number of scanners ASML can ship. It is the number of sub-suppliers who can qualify at all. Zeiss produces the mirrors at a single facility in Oberkochen, Germany, using ion-beam figuring and interferometric testing that takes months per optic. Cymer's laser-produced plasma light sources fire droplets of tin at 50,000 times per second, vaporising each with a CO2 laser to generate plasma that emits 13.5-nm light. The droplet generator alone is a proprietary electromechanical assembly with no publicly documented second source. TRUMPF supplies the high-power CO2 drive laser, built in Ditzingen, Germany. None of these suppliers can scale overnight, and none has a meaningful competitor.

The Shanghai Institute of Optics and Fine Mechanics, the Harbin Institute of Technology, and Huawei's various talent-recruitment programmes are all working on pieces of this supply chain from the Chinese side. The Diplomat's reporting makes clear that progress is real but uneven: the light source is further along than the optics, and the optics are further along than the resists. Even when each component reaches minimum viable performance, system integration remains. EUV lithography is not a sum of parts. It is a system in which the light source, the optics, the wafer stage, the reticle stage, and the resist chemistry all interact under vacuum with nanometre-scale tolerances, and where one weak link collapses throughput. China's 100-to-150-watt light source, 65-percent-reflective mirrors, and non-existent EUV photoresist supply represent three weak links operating simultaneously.

The equipment supply chain's forward trajectory will be shaped by three dates. ASML expects to ship 20 High-NA EUV systems in 2027, up from single digits in 2026. TSMC's A14 node, its first to require High-NA for critical layers, is scheduled for risk production in the first half of 2028. And China's 15th Five-Year Plan, which runs from 2026 to 2030, has designated "core lithography subsystems" as a national science and technology priority. The question is not whether each programme advances. It is whether any of them advances faster than the others bottleneck.