The EUV Materials War Behind AI Chips. Original analysis Not investment advice

EUV lithography is usually treated as an ASML story.

That framing is understandable. ASML builds the scanner. ASML creates the 13.5 nm light. ASML controls one of the most complex machines in semiconductor manufacturing. But the scanner is not the whole process. A wafer does not become an advanced chip just because EUV light hits it. The wafer has to be coated with photoresist, baked, exposed, baked again, developed, etched, cleaned, inspected, and integrated into dozens of other process steps. Every one of those steps can create defects, rough edges, bridge failures, missing holes, collapsed lines, or yield loss.

The EUV materials war is about turning scarce photons into working chips. It is the hidden half of every advanced wafer the AI industry consumes.

Photons are expensive.

---

Section 01 What the 2021 article got right

The 2021 SemiAnalysis piece on EUV photoresist, coaters and developers is the historical anchor for this essay^[1]. It argued that EUV had moved the semiconductor industry forward but had created new adjacent bottlenecks in masks, pellicles, deposition, etch, hard masks, and photoresist. Tokyo Electron was described as the dominant supplier of EUV photoresist coaters and developers. Japanese firms held roughly 75% of the photoresist market. JSR and Tokyo Ohka Kogyo were identified as the leaders in chemically amplified EUV photoresist. Lam Research was framed as trying to disrupt the stack with dry resist.

The article also explained the physics that make this a real battle. EUV has a throughput problem because far fewer photons hit the wafer compared with DUV at the same dose. Dose, sensitivity, roughness, resolution, and throughput are tightly linked. The lithography flow described on page 3 covers apply photoresist, prebake, align mask and expose, develop, then etch or implant. The stochastic-noise visual on page 6 shows the dominant defect modes at EUV pitches: line bridging, pinching, missing holes, and kissing defects^[1].

That framing was right. The EUV materials war was already a real battle in 2021. The 2026 question is how it evolved as AI demand pushed advanced wafer needs higher and tighter.

---

Section 02 Why EUV materials matter

EUV photons are hard to generate. Every EUV exposure needs enough photons to trigger a clean chemical or material response in the resist. More dose usually means better pattern fidelity, but lower scanner throughput. Less dose improves throughput, but increases stochastic defects. The resist has to balance sensitivity, resolution, line-edge roughness, defectivity, etch resistance, development behavior, cost, and high-volume manufacturability all at once.

The textbook way to summarize this is the RLS tradeoff: resolution, line-edge roughness, and sensitivity, plus the throughput and yield consequences that follow. Improving one of those terms usually puts pressure on another. The whole job of an EUV materials and process module is to find the point on the curve where the wafer is both manufacturable and economic.

That is the core economic tradeoff. Now apply it to three different materials philosophies competing for the same module.

---

Section 03 CAR, the incumbent

CAR means chemically amplified resist. It is the mature wet-resist path. The wafer is coated with liquid resist using spin coating, baked, exposed to EUV light, baked again to drive the chemical amplification, and then developed with a wet developer. CAR is not perfect, but it is deeply embedded in fab process flows.

CAR's advantages are operational. It is mature, well understood, qualified in production at scale, compatible with the existing TEL track tools that fabs already own, and supported by a large supplier ecosystem with familiar failure modes. Its weaknesses are physical. Stochastic defects, line-edge roughness, sensitivity tradeoffs, line collapse at thin films, wet-development residue, and harder scaling at the tightest pitches all get worse as EUV pushes deeper into the leading edge^[1].

CAR survives because fabs value what already works. But EUV has exposed its limits in ways that make process changes harder to defer indefinitely.

---

Section 04 MOR, the evolutionary challenger

MOR means metal oxide resist. It keeps much of the wet-resist process flow but changes the chemistry. Metal oxide materials can improve EUV absorption, which directly attacks the photon-scarcity problem and can help the RLS tradeoff. Inpria pioneered MOR and was acquired by JSR.

MOR is attractive because it can improve EUV sensitivity without forcing a fab to abandon the wet-track architecture. It still uses wet coating and development flows, so the wet-process problems do not disappear, but the path to qualification is shorter than a complete process-module change. The 2021 article also flagged that MOR can be used with TEL Clean Track tools with upgrades, and pointed to post-exposure bake and wet-development improvements from TEL and JSR. JSR and Inpria remain central to the MOR ecosystem, with JSR expanding photoresist development and production infrastructure in Japan and Korea^[7].

MOR is the evolutionary path. Better EUV absorption without forcing fabs to abandon the wet-track architecture they already trust.

---

Section 05 Dry resist, Lam's disruptive path

Lam's dry resist is the disruptive path. Instead of spin-coating liquid resist, Lam uses a vapor-deposited dry resist process. Instead of wet development, it uses dry development. The 2021 article showed Lam's case for fewer tradeoffs, higher resolution, wider process window, higher purity, and dry development that avoids the line collapse seen in wet flows^[1].

Lam's 2025 update put numbers on the lab progress. Lam said it had established 28 nm pitch high-resolution patterning through dry photoresist technology at imec, paired with low-NA EUV and described as extendible to High-NA EUV, with dry resist enhancing EUV sensitivity and resolution per wafer pass while using less energy and roughly five to ten times less raw material than existing wet chemical resist processes^[2]. In March 2026, IBM and Lam announced a collaboration on sub-1 nm logic scaling that explicitly covers new materials, advanced processes, and High-NA EUV lithography techniques^[3].

ASML prints the pattern, but materials determine whether the pattern survives.

Lam is not just trying to sell a resist. It is trying to own more of the EUV patterning module: deposition, dry resist, dry development, etch, and pattern transfer, treated as one co-optimized flow rather than a string of best-effort steps.

---

Section 06 TEL is defending the wet-track ecosystem

The wrong conclusion is "Lam disrupts, TEL dies." TEL is the incumbent in coaters and developers for a reason. Fabs value stability, throughput, installed base, process knowledge, and defect control. The wet-track ecosystem is not a single product. It is decades of process recipes, tool maturity, and qualification data, all of which matter when the cost of a yield excursion is large.

In December 2025, TEL released CLEAN TRACK LITHIUS Pro DICE, a 300 mm wafer coater and developer described as suitable for advanced lithography including High-NA EUV and DUV processes^[4]. In April 2026, TEL said the LITHIUS Pro DICE can reduce wafer defects and associated costs caused by flawed resist coating by 50% or more compared with previous models^[5]. TEL's LITHIUS product line continues to emphasize extensibility to advanced processes, high throughput, footprint reduction, OEE improvement, and cost-of-ownership reduction^[6].

Dry resist has to beat not just wet-resist chemistry, but decades of wet-track process maturity. That is a different and harder competition than chemistry alone.

---

Section 07 Why this became more important by 2026

The macro context made the materials war matter more. ASML reported strong Q4 2025 results, with Q4 net bookings of EUR 13.2B and EUV bookings of EUR 7.4B, against a year-end 2025 backlog of EUR 38.8B, with AI-related chipmaking demand driving the medium-term capacity picture^[10]. On the customer side, TSMC's N2 started volume production in Q4 2025 using first-generation nanosheet transistors^[11], and TSMC's A16 added Super Power Rail backside power delivery on top of nanosheets^[12].

More EUV scanners help. But they do not solve dose, defectivity, or pattern transfer on their own. If resist dose is high, scanner throughput falls. If stochastic defects rise, yield falls. If development collapses features, the wafer fails. If etch transfer damages features, the printed pattern does not become a working device. Buying more EUV scanners is not enough if the pattern does not survive.

Buying more EUV scanners is not enough if the pattern does not survive.

---

Section 08 The process-module war

The honest version of the materials war is that it is not just chemistry. It is full module optimization. Underlayer. Resist deposition. Coating or dry deposition. Post-apply bake. Exposure. Post-exposure bake. Development. Descum or residue removal. Etch. Clean. Inspection. Defect review. Every one of those steps influences yield, throughput, and cost.

TEL and JSR/Inpria want to extend and improve the wet process module. Lam wants to shift the module toward dry deposition, dry development, and tighter integration with etch. ASML provides the photon source, but materials and process tools determine how efficiently those photons become yield. This is not chemistry alone. It is process control.

---

Section 09 Why cross-licensing matters

The cleanest signal that the industry knows this is the recent wave of cross-licensing deals. In September 2025, Lam and JSR/Inpria entered a cross-licensing and collaboration agreement covering next-generation patterning, dry resist for EUV lithography, and advanced materials for atomic-layer etching and deposition^[8]. In 2026, Entegris and JSR/Inpria announced a non-exclusive cross-license covering metal oxide resist patents and collaboration on future photoresist materials^[9].

Competitors are also collaborators because advanced patterning is too complex for one company's chemistry alone. The ecosystem is converging around co-optimized flows, not simple winner-takes-all chemistry. The cross-licenses show the truth: the industry needs combinations, not slogans.

---

Section 10 CAR vs MOR vs dry resist, the likely 2026 verdict

The honest 2026 reading is that CAR, MOR, and dry resist are not competing for a single trophy. They are competing for layers. CAR remains where it is good enough and deeply qualified. MOR is the evolutionary path for wet EUV resist, especially where better absorption improves dose and defectivity without forcing a new process architecture. Dry resist is the disruptive path for the tightest, most photon-starved, and most defect-sensitive layers, especially as High-NA EUV moves closer to production.

The likely outcome is not one resist everywhere. It is layer-specific adoption tied to the dose, defectivity, pitch, etch transfer, and collapse risk of each layer in each customer's process.

---

Section 11 What people got wrong in 2021

The weak interpretation of the 2021 article was that Lam's dry resist would simply replace wet resist. That collapses the materials story into a chemistry-only story and ignores how fabs adopt new patterning flows. The better interpretation is that Lam identified a real weakness in wet EUV patterning, and that fabs will adopt new resist flows layer by layer, based on yield, cost, integration risk, and process control.

Semiconductor fabs are conservative because bad process changes destroy yield. A better lab result is not enough. The new flow has to survive high-volume manufacturing, defect inspection, tool uptime, contamination control, and customer qualification. In lithography, the best idea does not win until it becomes the best process of record.

In lithography, the best idea does not win until it becomes the best process of record.

---

Section 12 Risks and limits

The argument above blends the 2021 SemiAnalysis frame, company materials, and high-level macro context. It is worth being explicit about where the case can break.

The point is not that one material has already won. The point is that EUV scaling now depends on materials and process integration as much as scanner resolution.

---

Section 13 Final verdict

The 2021 article was right that EUV created a hidden battle around photoresist, coaters, developers, and etch. In 2026, that battle is more important because AI demand is pushing more advanced wafers, tighter pitches, and more expensive lithography capacity. CAR remains embedded. MOR is the evolutionary wet-resist path. Dry resist is the disruptive process-module path. TEL is defending wet tracks. Lam is trying to expand into the EUV module. JSR/Inpria sits at the center of MOR and the IP that underwrites it. ASML still supplies the photons, but the materials stack determines how many photons become working chips.

ASML supplies the photons. The EUV materials stack decides how many of those photons become working chips.

---

Section 14 Evidence ledger and source notes