Author: Li Yan
The rapid advancement of AI (Artificial Intelligence) is driving a continuous surge in global computing power demand. Meanwhile, energy supply is gradually becoming a major constraint on the expansion of computing infrastructure. The International Energy Agency (IEA) estimates that by 2030, global data center electricity consumption will nearly double compared to current levels, with AI-related data centers showing particularly significant growth in power demand. Consequently, the technical value of advanced nodes is reflected not only in the continuous miniaturization of transistor sizes and the improvement of computing performance but also in steadily improving computing performance and energy utilization efficiency under limited power constraints. Therefore, advanced nodes supported by EUV (Extreme Ultraviolet) Lithography and High-NA EUV have become critical underlying technologies for continuing semiconductor scaling and improving the performance-per-watt of high-performance computing chips.
However, this energy efficiency challenge exists not only in the usage phase of chips but also throughout their manufacturing process. imec research shows that in advanced logic chip manufacturing, lithography and dry etching combined can account for nearly 40% of carbon emissions during the manufacturing phase, and the energy consumption of EUV lithography equipment is significantly higher than that of traditional 193 nm immersion lithography equipment. One important reason is the complex energy conversion process of the EUV light source: a high-power CO2 laser bombards tin droplets at a frequency of tens of thousands of times per second, turning them into plasma and generating EUV radiation with a wavelength of 13.5 nm, which is then completed for exposure through a multi-stage reflective optical system.
In 2025, ASML experimentally achieved an EUV light source power of approximately 1,000 W and increased the tin droplet interaction frequency to 100 kHz, meaning about 100,000 laser-tin droplet interactions per second. As a result, the R&D goals for advanced EUV systems are no longer just pursuing higher resolution and light source power, but achieving system optimization among light source conversion efficiency, exposure dose, wafer throughput, equipment power consumption, and energy consumption per wafer. How many effective EUV photons can be generated and how many wafers can be exposed per 1 MWh of input power, ultimately translating into how much advanced computing power—even if the EUV generation efficiency improves by only a few percentage points, under conditions of large-scale installation and long-term high utilization rates, amplified by the global installed base, it could translate into tens or even hundreds of GWh of lifecycle energy savings. Therefore, the competition in advanced lithography technology is expanding from mere dimensional scaling to the collaborative optimization of resolution, production efficiency, and energy efficiency.
At this juncture, Lin Nan, a light source expert who has worked within the ASML system for many years, returned to China and currently serves as a professor at Beihang University. His research directions include EUV light sources, chip metrology and inspection, and the generation and application of attosecond pulses. In recent years, Professor Lin Nan founded Huaxin Jiguang and has been committed to improving EUV light source technology and enhancing conversion efficiency. Recently, Semiconductor Industry Perspectives conducted an exclusive interview with Professor Lin Nan, engaging in an in-depth discussion on his team's route selection and technological breakthroughs focusing on the lithography industry.
01. Technological Route Debate and Strategic Choices
Regarding where China's lithography technology is heading, there are two typical judgments in the industry. One view holds that external restrictions will force China to take a completely parallel technological route, developing and building everything from the whole machine to components independently; another view points out that lithography is one of the most complex system engineering projects in human manufacturing history, involving top-tier accumulations in multiple disciplines such as optics, precision machinery, materials, and control. No single economy can independently complete the entire puzzle.
Lin Nan believes, "China should take a pragmatic route of limited parallel development, developing core components and materials in parallel, which is unquestionable; but for common technologies, we should have an international vision to support the evolution of next-generation technologies." In his view, the scale of China's chip industry is already large enough, and the key to "becoming strong" lies in innovation, which not only solves its own problems but also possesses the ability to provide solutions for the global supply chain.
In the field of lithography, the light source is a key technological point. EUV lithography uses extreme ultraviolet light (EUV) with a wavelength of 13.5 nm. Currently, there are several technical paths for light source generation: Laser-Produced Plasma (LPP), Discharge-Produced Plasma (DPP), Laser-Assisted Discharge Plasma, and High-Harmonic Generation (HHG). However, there is only one mainstream solution for mass production—LPP. The principle of the current commercial LPP EUV light source is that tens of thousands of molten tin droplets are ejected from the generator per second. The pulsed laser first uses a pre-pulse to flatten the tin droplets into a pancake shape, and then the main pulse bombards them into a plasma state of hundreds of thousands of degrees, radiating the characteristic spectrum of 13.5 nm; the multi-layer mirror collects these lights laterally and sends them into the illumination system.
Currently, only the Netherlands' ASML has achieved mass production of this technology. The power, stability, and energy consumption of the light source directly determine the capacity ceiling of the lithography machine. The industry commonly uses "Wafers Per Hour" (WPH) to measure the value of a lithography machine, and the light source power is one of the keys. The technological route chosen by Lin Nan is to use a solid-state laser to drive Laser-Produced Plasma (LPP) to generate EUV light.
After returning to China, he did not start a business immediately but first joined research institutes and universities. The reason is that a new principle light source requires broad supply chain support. "I needed to return to China, in my own soil, to truly make this idea a reality first." With this in mind, he began five years of preliminary exploration and technical verification. In the laboratory, he gradually overcame the localization challenges of core technologies; more importantly, he found that China has already gathered a group of component suppliers with technical capabilities who are willing to invest together and tackle the hard nuts with the team. The convergence of technological breakthroughs and the maturation of the local supply chain ultimately led to the birth of Huaxin Jiguang.
02. Redefining EUV Light Source Efficiency
To evaluate the pros and cons of EUV light sources, the core indicator long used by the industry is Conversion Efficiency (CE), which is the proportion of incident laser energy converted into EUV energy. This indicator is certainly important, but it only describes half of the chain; it does not account for the electrical energy consumed to generate these lasers. However, while the industry has optimized the plasma stage, it has long ignored the huge loss space at the laser end.
Lin Nan advocates replacing it with system efficiency: multiplying the electro-optical efficiency of the laser by the conversion efficiency of the plasma to obtain the complete chain efficiency from electricity to EUV light. This redefinition is of great engineering significance. It means there are two multiplicative variables for system optimization. The traditional approach only pushes hard on the plasma end, leaving the order-of-magnitude improvement space at the laser end idle.
From the practical results, the numbers also confirm the feasibility of this path. The plasma generation efficiency of commercial EUV lithography machines has reached over 5.5%, while the efficiency of the laboratory prototype of Lin Nan's team has reached 4.4%, equivalent to 80% of the commercial level; the gap lies in the laser end. In the first generation of mass-produced EUV lithography machines, the efficiency of the CO2 gas laser was only about 2%. By adopting a new architecture solid-state laser, the electro-optical efficiency of the laser can easily reach over 10% or even 20%. With the rise of one and the fall of the other, the overall efficiency surpasses that of commercial light sources. It is understood that Lin Nan's team uses an LPP EUV light source driven by a solid-state laser operating in the near-infrared or mid-infrared band, achieving an electricity-to-EUV conversion efficiency of more than three times that of mass-produced EUV light sources.
Of course, good data does not mean it can enter the production line. Lin Nan also admitted that the current comparison is based on single-pulse verification in the laboratory—one laser pulse to one EUV pulse—while true engineering products need to increase the pulse repetition frequency to the level required for mass production, which requires a massive investment in engineering. This is exactly why he established an industrialization company and introduced market-oriented funding. From the laboratory to stable mass production, there is a long engineering chasm.
03. Reflections on the Path to Industrialization
How to cross the threshold from the laboratory to industrialization, Huaxin Jiguang always considers technological innovation and substitution from the user's perspective.
In terms of industrialization strategy, it is not about starting from scratch, but replacing the core without changing the housing. Lin Nan stated that for equipment manufacturers, the biggest concern about replacing core components is system compatibility and verification costs. Therefore, Huaxin Jiguang's R&D is strictly constrained under the premise that "all key external interface parameters of the light source remain unchanged," only realizing the principle replacement inside the light source. For downstream customers, the replacement action is nearly plug-and-play, and the migration cost is almost zero; while in operation, the differences in efficiency, power consumption, and maintenance costs will naturally emerge. Allowing customers to complete the technological upgrade with almost no perception is the underlying logic for this route to be quickly accepted. After all, energy saving is a common demand. Customers do not need to make a stance choice between whether to use new technology or not, and can naturally accept the new light source technology route.
In terms of industrialization entry, choose to walk on two legs. First, use EUV metrology and inspection light sources to meet current market demand and form a self-sustaining cash flow, then feed back into R&D to support the long-term vision of future EUV lithography light sources. Regarding starting with EUV metrology and inspection light sources, Lin Nan further explained: metrology and inspection equipment in the front-end and packaging processes combined account for about 20% of the global semiconductor manufacturing supply chain, with a scale comparable to lithography equipment. But unlike lithography where one machine solves one type of problem, metrology and inspection involve a large number of key parameters, a wide variety of equipment, and follow the rule of one generation of process, one generation of equipment. Every time a new process indicator appears, a brand-new measurement equipment needs to be matched. Lin Nan stated that the current period is exactly the transition node of the metrology and inspection indicator system. At the 7nm node and above, overlay accuracy (overlay, measuring the alignment degree of the front and back two layers of patterns) and Critical Dimension (CD, measuring the line width control accuracy) are two independent measurement objects; entering the sub-2nm node, the relative position error between the two layers of patterns can no longer be split into independent variables and must be measured as a whole—this is Edge Placement Error (EPE). Around EPE measurement, a whole set of metrology and inspection demands that did not exist before will be spawned. At the same time, there are differences in the IC (Integrated Circuit) process path between China and overseas, which will also derive additional measurement demands that the overseas market does not have. Moreover, metrology and inspection equipment has relatively low requirements for light source power and stability, and a short verification cycle, making it the best training ground for engineering.
In terms of industrialization implementation, what is truly difficult is the accumulation of customer trust. Lin Nan emphasized that although your parameters are indeed better than others; this alone is far from enough. "There are too many companies in the semiconductor industry that have achieved better individual parameters than competitors, but customers dare not use them because they have not been tested by long-term stable operation." In addition, it is also necessary to face whether customers are willing to open key data to you and whether they are willing to bear the risk of iterating with you. Semiconductor process data is core confidential information, and trust can only be accumulated through delivery after delivery. He mentioned that his experience at ASML made him understand that long-term cooperative customers are willing to bear this risk because they believe you have the ability to solve their problems. But opening up trust is not that easy.
Going global also requires crossing another threshold: patents. Half of the competitive barriers in the lithography field lie in technology, and the other half in patents. ASML and its upstream supply chain have established a dense patent network around the LPP light source driven by CO2 lasers, which is a minefield that latecomers cannot bypass or afford to bypass. Based on the new technological route, Lin Nan personally owns more than 120 international patents. Lin Nan stated, "Our current technical solution has no conflict with the existing patent layout in any region of the world." This route can not only serve the Chinese market but also enter the global supply chain, becoming a supplementary option outside the ASML system. Lin Nan's new technological route is not about closing the door to rebuild a system, but using China's complete industrial chain resources and huge application market to cultivate a technical solution that can be exported globally, becoming a new option in the global supply chain.
04. 2026: Where is Lithography Heading?
Bringing the time coordinate to 2026, the global lithography industry is standing at a critical juncture.
EUV lithography has entered a comprehensive mass production period. Since TSMC (Taiwan Semiconductor Manufacturing Company) launched the first 7nm EUV chip in 2019, the most advanced 2nm node has been mass-produced at Samsung and TSMC. And the industry's focus is turning to High NA (High Numerical Aperture) models. ASML's 0.55 NA lithography machine has had 10 units delivered and put into use by 4 customers, with another 3 units in transit. High NA models replace multiple exposures with a single exposure, which is the ticket to sub-1nm nodes. At the same time, Hyper-NA EUV lithography machines with larger numerical apertures are also under development.
The debate over the technological route of light sources is also heating up. Currently, there are multiple technical routes for EUV light sources: Laser-Produced Plasma (LPP), Discharge-Produced Plasma (DPP), Free Electron Laser (FEL), Synchrotron Radiation (SR), High-Harmonic Generation (HHG), etc. Lin Nan's attitude is want them all scientifically, but make choices commercially: all technical routes should be encouraged and tried; but which routes can be commercialized faster and are more worthy of capital investment to accelerate requires prioritizing based on the judgment of technological maturity. Currently, the 6.x nm wavelength BEUV (Beyond EUV) light source has been put on the agenda by the academic community. The 0.75 NA architecture is sufficient to support the industrialization of the next decade or so, which leaves a comfortable R&D window for the next-generation light source. Lin Nan's expectation is: "Is it possible for Chinese scientific researchers to leverage the advantages of complete industrial supporting and rich application scenarios in the R&D of next-generation light sources and propose their own solutions?"
The industry is doing 5 to 10 years of research, while the academic community is looking 10 to 20 years into the future. Lin Nan happens to stand at the intersection of these two timelines. Behind him is the deep accumulation of five years of laboratory verification, and in front of him is the long road from metrology and inspection to lithography, from domestic to global. Returning from overseas to China's industrial soil, turning the light in the laboratory into a light source that illuminates the production line, in his view, this is a very exciting moment. Being able to practically solve industrialization problems through technological accumulation in such a wave is also what the times require. With a long slope and thick snow, we look forward to Professor Lin Nan and his team welcoming the day of industrialization and scaling as soon as possible!