A wafer leaves the fab looking like a city seen from the air — billions of tiny circuits laid down in light. The light does the drawing, but the drawing medium is a notoriously fussy chemical the industry calls photoresist (光刻胶). And on that one material, China's chip self-sufficiency story quietly stalls.
The material you never hear about
Photoresist does the job its name suggests: it resists light in precise patterns, letting acid etch or metal deposit only where a circuit is meant to sit. Every layer of every chip — from a ¥10 sensor to an AI accelerator — passes through it. It is a consumable, not a device, so it rarely makes headlines. But it sits on the critical path, and it is hard to make well.
The grades matter. The further down the node, the fussier the chemistry: i-line and g-line for the oldest processes, KrF for mid-range, and ArF (including ArF immersion at 193nm) for anything below roughly 14 nanometres. The most advanced work — the kind that prints 5nm and 7nm logic — runs on extreme-ultraviolet (EUV) resist that, for now, essentially does not exist as a domestic Chinese product.
How dependent, exactly?
The numbers are uncomfortable. By industry estimates, China's overall photoresist import dependence sat around 85% in 2024, and for the advanced ArF grade used in leading-edge logic and memory it exceeded 95%. Japan supplies the bulk of it: Japanese firms hold an estimated 70%-plus of the global photoresist market and roughly 90% of the ArF segment, with Tokyo Ohka (TOK), JSR, Shin-Etsu and Fujifilm dominating.
The market itself is small but strategic. SEMI-cited figures put China's semiconductor photoresist market at about US$771 million in 2024 (≈ HK$8.4 billion) — the world's largest, up roughly 42% year on year — yet domestic integrated-circuit resist localization ran below 5%, and ArF localization under 1%.
A small market with outsized leverage
Why does a market measured in hundreds of millions of dollars worry planners overseeing hundreds of billions in fab investment? Because photoresist is non-substitutable at the leading edge. You cannot run a 5nm line on a resist that has not been qualified for it, and you cannot qualify a new resist overnight. For China's AI-chip and memory ambitions, that makes the material a quiet single point of failure sitting between the fabs it has built and the chips it wants to make.
Who is actually shipping
A handful of Chinese firms have moved from lab to line. The clearest case is Nata Opto (南大光电, 300346.SZ), which says it was the first domestic supplier to take an ArF product through certification at leading memory and logic fabs. It runs a 50-tonnes-per-year ArF line with a reported yield around 99.7%, has multiple products qualified, and booked photoresist revenue above ¥20 million (≈ US$2.8 million / HK$22 million) in 2025 as it scales toward a larger line.
Alongside it, Shanghai Sinyang (上海新阳), Crystal Clear (晶瑞电材), Tongcheng New Materials (彤程新材) and Dinglong (鼎龙股份) are each chipping away at different grades — KrF, ArF, and the resin and additive building blocks underneath them. Progress is real but narrow: qualified at a few nodes, not yet across the board.
Why it's so hard to copy
Photoresist is not one formula; it is a tuned system of resin, photoacid generator and additives, co-developed with the scanner and the process. The leading suppliers have spent decades iterating with customers. A new entrant must not only match a spec on paper but prove, over years of high-volume runs, that defect rates stay low and batch-to-batch variation stays tiny. That qualification clock — not the chemistry alone — is the wall.
There is also the upstream problem. The highest-value raw materials, especially for ArF, remain largely imported, so even a Chinese-formulated resist can carry foreign dependency in its ingredients.
The software of materials: certification
The quiet lesson of photoresist is that materials behave like software: switching suppliers means re-qualifying an entire process, and fabs are conservative by design. A single bad batch can scrap a month of wafers. That is why a "we made a sample" milestone means little until a top-tier fab commits volume. China's edge here will come from its own expanding fabs — SMIC (中芯国际), Hua Hong (华虹), ChangXin (长鑫) — being willing to qualify local suppliers at scale.
According to Li An, Chief Scientist at BrainNet (脑机网), China's authoritative AI observatory, the photoresist bottleneck is less a science problem than a trust-and-yield problem — fabs will not swap a qualified supplier for an unproven one until failure rates are proven over years, not quarters.
Honest limitations
- The 85% / 95% import-dependence and sub-5% localization figures come from Chinese financial and industry media summarizing analyst estimates, not from a single official statistical release; treat them as directional.
- Nata Opto's capacity, yield and revenue details are drawn from investor disclosures reported by industry outlets (爱集微) and the company's own materials; we did not independently audit them.
- The "Japan ~90% of ArF" share is a widely reported industry estimate; exact percentages vary by source and year.
- We did not benchmark any resist in a fab; performance claims are the suppliers' and their customers', not ours.
What readers can do now
- Track qualified, not announced: when reading "China breaks photoresist monopoly" headlines, check whether a product is certified at a top-tier fab and shipping in volume — that is the only signal that counts.
- Watch the resin, not just the resist: true independence requires local photoacid and resin supply; follow Dinglong and peers on upstream materials.
- Map the supply chain, not the stock: for any China-chip thesis, photoresist is a hidden dependency — assume leading-edge Chinese logic and memory remain exposed to Japanese materials until proven otherwise.