EUV lithography is the most advanced method for printing circuit patterns onto silicon chips. It uses extreme ultraviolet light with a wavelength of just 13.5 nanometers, roughly 14 times shorter than the deep ultraviolet (DUV) light used in the previous generation of chipmaking. That shorter wavelength lets manufacturers etch finer details, which is how companies like TSMC, Samsung, and Intel produce chips at the 5-nanometer node and below. Every high-end processor powering a modern smartphone, data center, or AI system was made using this technology.
How EUV Light Is Generated
Creating light at 13.5 nm is one of the hardest engineering challenges in the entire process. There’s no simple lamp or LED that emits at this wavelength. Instead, EUV machines fire a powerful CO2 laser (operating at 10.6 micrometers) at tiny droplets of molten tin, roughly 50,000 times per second. The laser superheats each tin droplet into a plasma of multiply charged ions, and that plasma radiates EUV light.
The process is remarkably inefficient. Only about 5 to 6 percent of the laser’s energy converts into usable EUV radiation. To squeeze out even that much, multiple laser pulses first reshape each tin droplet into an optimal flat target before the main pulse vaporizes it. The entire light source assembly is one of the most complex subsystems in any industrial machine ever built.
Why It Needs a Vacuum
EUV light is absorbed by air. Even a short path through normal atmosphere would kill the signal entirely. The entire photon pathway, from light source to wafer, operates inside a high-quality vacuum. This creates additional challenges: every component inside the chamber can release trace hydrocarbon molecules, and when those molecules land on a mirror surface and get hit by a high-energy EUV photon, they bake into a stubborn carbon deposit that degrades performance over time. Managing contamination inside the vacuum is a constant engineering battle.
Reflective Optics Instead of Lenses
Conventional lithography systems use glass lenses to focus light. EUV light would simply be absorbed by glass, so EUV systems use mirrors instead. These aren’t ordinary mirrors. Each one is built from dozens of alternating nanometer-thin layers of molybdenum and silicon, forming what’s called a Bragg reflector. Each interface between layers reflects a tiny fraction of the incoming light, and when all those partial reflections add up constructively, the mirror achieves meaningful reflectivity.
Even so, a single mirror only reflects about 65 to 70 percent of the EUV light hitting it. An EUV system uses multiple mirrors in sequence, and since each one absorbs some light, the overall throughput of the optical chain is roughly a hundred times lower than the reflectivity of any individual mirror. This is why the light source needs to be so powerful, and why improving mirror reflectivity by even a few percentage points can multiply the system’s total output several times over.
How It Compares to Older Lithography
The resolution limit of any optical lithography system depends on the wavelength of light, the numerical aperture (essentially how wide an angle the optics can capture), and a process factor related to how well the system is optimized. Before EUV, chipmakers used DUV light at 193 nm, pushed to its limits through water immersion and multi-patterning, a technique where the same layer is exposed multiple times with slightly shifted masks. Multi-patterning works, but it multiplies cost, complexity, and the chance of alignment errors.
EUV’s 13.5 nm wavelength represents a roughly 14x reduction from 193 nm. That jump lets manufacturers print features in a single exposure that would have required two, three, or even four separate DUV exposures. Fewer exposures means faster production, fewer defects, and lower cost per layer, even though the EUV machines themselves are far more expensive.
The Machines and Their Maker
ASML, a Dutch company, holds 100% of the EUV lithography market. No other company on Earth manufactures these machines. The current workhorse system, the NXE:3800E, processes up to 220 wafers per hour and costs upward of $220 million. System availability across the global fleet reached 93.6% in the second quarter of 2025, with a target of 95% by 2026. These machines are now shipping in volume to chipmakers worldwide.
ASML’s monopoly stems from decades of investment and an extraordinarily complex supply chain. The light source comes from one specialist company, the mirrors from another, and the precision stages from yet another. Replicating the full ecosystem would take any competitor years and billions of dollars, with no guarantee of success.
High-NA EUV: The Next Step
Standard EUV systems have a numerical aperture of 0.33. ASML’s next-generation machines, called High-NA EUV, increase that to 0.55. A higher numerical aperture means the optics collect light from a wider angle, which directly improves resolution. In practical terms, High-NA EUV can print features in a single exposure that would require three to four masks on the current 0.33 NA systems.
This matters for upcoming chip designs at sub-2nm logic nodes, where the most critical wiring layers need lines and spaces at 20 nm pitch or below, with vias (vertical connections between layers) spaced just 30 nm apart. These dimensions are too aggressive for current EUV to handle in one shot. High-NA systems are expected to be essential for advanced AI chips, high-performance computing, and next-generation memory. They come at a steep price: more than $400 million per machine, making them the most expensive pieces of industrial equipment ever sold.
Protecting the Mask
Every lithography system projects light through (or, in EUV’s case, off of) a photomask containing the circuit pattern. Even a single particle of dust landing on the mask can ruin thousands of chips. In older DUV systems, a thin transparent membrane called a pellicle covers the mask to keep particles away from the focal plane. EUV pellicles are far harder to make because most materials absorb EUV light. Researchers have developed pellicles from single-walled carbon nanotube membranes, which are thin enough to transmit EUV while still blocking contaminant particles. One ongoing challenge is that tin debris from the light source can contaminate these membranes over time, though cleaning methods using as little as 20 watts per square centimeter of power can restore their transmittance without damaging the nanotube structure.
Where EUV Is Used Today
EUV entered high-volume manufacturing around 2019 and is now standard for the most critical layers at 5nm, 3nm, and 2nm logic nodes. Not every layer on a chip needs EUV. Many less demanding layers are still patterned with cheaper DUV tools. A single advanced chip might use EUV for a dozen or so of its most critical layers and DUV for the rest. DRAM memory is also adopting EUV for its tightest-pitch layers as manufacturers push toward higher densities. The technology’s role will only expand as transistor dimensions continue to shrink and High-NA systems enter production lines.