in short
Global neon consumption for chipmaking was only about 540 metric tons in 2021. Qualifying a new photoresist at the leading edge can take up to three years. And China's export licensing on gallium, rare earths and at least seven other materials has been tightening, layer by layer, since 2023. The chip industry has now run the same experiment three times on three different materials, and the results are very different. One taught the industry to diversify. One it has not learned from. And one has become a weapon.
The cornerstone piece in this series described materials as the second bottleneck between the world and unlimited AI compute: a handful of cheap, boring inputs that a fab cannot run without, and that almost nobody built redundancy for. That framing is right, but it hides something more useful. Neon, photoresist and the critical minerals are three different stages of the same story, and comparing them shows what actually makes a materials bottleneck fixable or permanent. The answer has less to do with how scarce the material is than with how hard it is to replace once a fab has been tuned around it.
Why Materials Are a Different Kind of Bottleneck
Machines, people and water are all expensive, visible and slow to expand. Materials are the opposite. They cost almost nothing relative to the factory that depends on them, they arrive quietly in cylinders and drums, and no executive thinks about them until the day a shipment doesn't show up. That mismatch between cost and consequence is the whole problem. A material that is cheap enough to ignore is also cheap enough that nobody pays for a second supplier, and a second supplier is exactly what you can't improvise during a crisis.
There is a second feature that matters even more: a fab doesn't just use these materials, it is tuned to them. Every process step has been optimized around a specific gas purity, a specific chemical formulation, a specific supplier's quirks. Swapping an input isn't a procurement decision. It's a re-engineering project. How long that project takes is what separates the three stories below.
Neon: The Pilot Light
Think of a furnace's pilot light. It's a tiny flame that costs almost nothing to run, and when it goes out, the entire furnace is dead. Neon plays that role in chipmaking. It is a key ingredient in the gas mixtures that power the excimer lasers used in deep-ultraviolet lithography, and there's no workable stand-in for it: helium can't replace neon without degrading laser performance (SpecGas). The volumes are small, roughly 540 metric tons consumed globally for chip production in 2021 (Global News). The consequences of running short are anything but.
The supply was historically concentrated in an odd place. Neon is a by-product of industrial air separation, and Ukraine's producers, operating near its major steelworks, ended up with an outsized share of chip-grade output, estimates of which run from about 50 percent to as high as 70 percent depending on the year and the source (Epignosis Insights; USITC; SpecGas). When Russia's full-scale invasion in 2022 halted Ukraine's two leading suppliers, prices surged by more than 500 percent (SpecGas).
The Warning the Industry Ignored Once
What makes the neon story instructive is that the 2022 shock wasn't the first. When Russia annexed Crimea in 2014, neon prices reportedly spiked by more than 600 percent, and Ukraine supplied around 70 percent of chipmakers' neon at the time (USITC). Some manufacturers diversified afterward, and Ukraine lost roughly a fifth of its market share over the following years. But the broader industry returned to business as usual, with no meaningful diversification investment in the eight years that followed (SpecGas). It took a second, larger shock to change behavior.
Why Neon Is the Good News
Here's the part the cornerstone only hinted at: after 2022, neon actually got fixed, or at least substantially fixed. China is now described as the largest active neon producer and commercial supplier in the world. Air Liquide brought a new ultra-high-purity neon purification unit online in Louisiana in February 2026, adding 200,000 cubic meters a year, and Linde secured a multi-year exclusive supply agreement with a major US chipmaker in January 2026 (SpecGas). Price trackers now show rare-gas costs easing from their peaks, though still well above pre-2022 levels, with Ukraine's capacity only partially restored and the recovery driven mostly by new capacity in China, South Korea and the United States (Epignosis Insights).
Why did neon diversify when other materials haven't? Because neon is, in the end, a commodity gas. Building a new purification plant takes capital and time, but a fab doesn't have to re-engineer anything to use neon from a different plant, so long as the purity meets spec. The substitution cost is low. That single property is what made diversification possible, and it's the property the next material doesn't have.
Photoresist: The Paper Every Darkroom Is Tuned To
Think of a photographic darkroom. The enlarger projects the image, but the photographic paper is what actually captures it. A professional darkroom spends years tuning every step, from exposure time to developer temperature, around one specific paper. Switching to a different brand isn't a quick swap. Every test print has to be redone. In chipmaking, the lithography machine is the enlarger, and photoresist is the paper: a light-sensitive coating on the wafer that records the pattern the machine prints.
The photoresist market is concentrated in a way that makes neon's pre-2022 concentration look mild. Four Japanese companies, Tokyo Ohka Kogyo, JSR, Shin-Etsu Chemical and Fujifilm, lead the market, and estimates of Japan's share of global photoresist run from about 70 percent to about 90 percent depending on the source (Intel Market Research; SEMI VISTA). For the extreme-ultraviolet resist used at the most advanced nodes, estimates run from over 85 percent to about 95 percent (Persistence Market Research; Asia Times). Even the ownership has a strategic cast: JSR, the market leader, was acquired by the Japan Investment Corporation, a government-backed fund, in a deal worth roughly $6.4 to $6.9 billion and delisted in 2024 (Public Markets).
You Can't Just Switch Suppliers
The reason this concentration persists is the darkroom problem. Only three suppliers, Tokyo Ohka Kogyo, JSR and Shin-Etsu, are qualified as volume suppliers of EUV resist at leading-edge foundries, and each qualification is specific to a foundry, a node and a set of process layers (SemiconductorX). Qualifying a new resist formulation for high-volume manufacturing involves lab evaluation, pilot-line trials, defectivity analysis and long-term stability testing, a process estimates put at 12 to 36 months for the most advanced nodes (Fortune Business Insights). For a new supplier trying to enter, the same industry analysis puts the timeline at three to five years and tens of millions of dollars in capital (IndexBox). Changing suppliers, in the words of one industry source, is not a procurement decision. It is a process re-qualification (SemiconductorX).
Compare that with neon. A neon plant is built once and then sells a commodity gas to anyone who meets purity specs. A photoresist supplier has to be married to a specific fab's process before it can sell anything at all. That's why photoresist remains concentrated years after everyone agreed it should be diversified.
When the Lock Became a Lever
For a long time, concentration like this was an abstract risk. It stopped being abstract in late 2025. According to one industry analysis, Japan's trade ministry placed twelve core semiconductor materials, including high-end ArF and EUV photoresists, on an export-control list that restricted supply to 42 named Chinese companies (Public Markets). Reports that Japanese suppliers had tightened raw-material supply and slowed equipment servicing circulated widely in Chinese media, and Beijing's response included a stated goal of supplying 40 percent of its own photoresist needs in 2026 (Asia Times).
The pressure isn't only political. In 2026, disruptions to naphtha supply, the petrochemical feedstock behind many resist solvents, led Japanese photoresist suppliers to warn major Korean chipmakers of potential shortages, since Japan relies on the Middle East for more than 40 percent of its naphtha (TrendForce; Mordor Intelligence). The chain reaches from a Gulf refinery to a Korean cleanroom through a few chemical plants in Japan, and none of those links has a spare.
China's attempt to build its own supply illustrates how long the road is. Domestic makers are adding capacity, with one reportedly bringing KrF and ArF resist production to 330 tons a year in 2026 (Mordor Intelligence). The top of the ladder is harder, because EUV resist can only be qualified on EUV machines. China has never been able to buy one. Reuters reported in December 2025 that a Chinese team had built a prototype EUV machine that can generate EUV light but has not yet produced working chips, with working chips targeted for between 2028 and 2030 (Engadget, summarizing Reuters; Power Electronics News). Until a machine like that works at production scale, there is no way to qualify an EUV resist in a production environment. The materials problem and the machine problem turn out to be the same problem.
Critical Minerals: The Mill, Not the Mine
Think of a wheat country with exactly one flour mill. The wheat can be grown anywhere. But if only one mill knows how to grind it to the required standard, whoever owns the mill controls the bread. The critical minerals story works the same way: the leverage sits in refining, not mining. China produces more than 60 percent of the world's rare earth mining output but over 85 percent of its processing capacity, according to USGS figures (Rare Earth Mining), and the processing step took decades of industrial know-how, environmental tolerance and low-margin patience that few other countries chose to build.
Gallium and germanium, essential to compound semiconductors and certain chip types, show the pattern even more starkly. Depending on whose count you use, China supplies somewhere between 80 and 99 percent of the world's gallium and a dominant share of its germanium (Metal Tech News). The Council on Foreign Relations counts export controls imposed between 2023 and 2026 on gallium, rare earths and at least seven other materials, extending to products made from them such as permanent magnets and semiconductor wafers (CFR). China's gallium and germanium exports to Japan dropped to zero in the early months of 2026, according to Chinese customs data (Metal Tech News).
What Export Controls Actually Do to Prices
The most telling evidence isn't the announcements. It's the prices. The International Energy Agency's 2026 critical minerals outlook found that prices for these minerals had more than doubled, with tungsten surging sixfold, and that export controls had split the market in two: gallium and heavy rare earths in Europe were trading at around five times Chinese domestic prices, and germanium at nearly three times (IEA). That divergence is the real price of a chokepoint. Buyers outside the controlled supply chain aren't paying more because the material got scarcer in the ground. They're paying more because access to it got conditional.
The stakes extend beyond chips. The IEA has estimated that a complete Chinese block on rare-earth permanent magnet supply alone would cost the United States and Europe combined around $3.5 trillion a year (CFR, citing IEA). And the controls are designed to follow the material downstream: under the rules announced in October 2025, an export license could be required for a product containing as little as 0.1 percent Chinese-origin rare earth content by value (Tech Insider). That package is currently suspended, which brings us to the dates that matter.
Not Every Lever Is Pulled at Once
The picture is not one of permanent blockade, and honesty requires saying so. China suspended its gallium, germanium and antimony ban on exports to the United States in November 2025, but the suspension is time-limited, runs through November 27, 2026, and leaves the underlying licensing requirement in place (Reuters, via Investing.com). The broader October 2025 rare-earth package, including the 0.1 percent rule, is suspended until November 10, and according to trackers no extension of either date has been announced, with China's commerce ministry saying in April that any arrangement after November is subject to US–China consultation (Materials Dispatch; AsiaAI tracker). The two sides have reportedly agreed to extend their wider trade truce to January 10, 2027, but it is unclear whether that covers these suspensions (AsiaAI tracker). That makes the current calm look like a pause rather than a settlement, with two deadlines inside the next seven weeks. Chinese commentators have also argued that the leverage runs both ways, since a Japanese photoresist embargo could be answered with rare-earth restrictions that Japan's magnet and vehicle industries would feel immediately (Asia Times). When two parties each hold a different chokepoint, the result is closer to a standoff than a monopoly, which can be stable, until it isn't.
Why Capital Doesn't Fix This Quickly
A subsidy can build a refinery. It can't compress the time it takes to build a skilled refining workforce, win an environmental permit, or qualify a replacement resist at a fab that has spent years tuning around the incumbent. The pattern across all three materials is the same: the cheaper and more boring the input, the less anyone invested in redundancy, and the longer the gap between deciding to diversify and being able to. The expense lands in the worst possible place, at the point where a disruption has already begun.
The Challenge: Markets Do Diversify When Pushed
The fair counter-argument is that neon proves markets respond. Within four years of the 2022 shock, new purification capacity opened in the United States, China and South Korea, and long-term supply contracts replaced spot buying. China is racing to build photoresist capacity, and Western governments are funding mines and refineries. Every prior materials crunch has eventually produced new supply, and treating today's chokepoints as permanent risks overlooking the same adaptive pressure that eventually solved neon.
The honest answer is that the three materials differ in how fast adaptation can happen. Neon needed new plants. Photoresist needs new plants and years of process qualification inside foundries that must agree to take the risk. Critical minerals need new plants, new expertise, new permits and, often, a price floor that makes Western refining economic against a competitor who can undercut it at will. The speed of the fix depends on the depth of the lock-in, and the lock-in is deepest exactly where the technology is newest.
Conclusion
Materials are the bottleneck that looks least like a bottleneck, because it is made of things that are cheap, small and easy to overlook. Neon showed that the lesson can be learned, after two crises and eight years of delay. Photoresist shows what happens when the lesson hasn't been learned yet and substitution takes years. The critical minerals show what happens when someone decides to use the chokepoint deliberately. The unit cost of each input is trivial. The cost of not having it is a factory.
This piece expands the second bottleneck in The Four Things Standing Between Us and Infinite AI Compute. Its companions cover the talent gap and the water problem.
Sources
source
covers
2014 neon shock; Ukraine's share; early diversification
Helium non-substitutability; 2022 price surge; industry inaction after 2014
China as largest producer; Air Liquide Louisiana; Linde agreement
Global chip-grade neon consumption (540 metric tons)
2026 neon price trend; partial Ukrainian recovery
Japanese suppliers' ~90% photoresist share
EUV resist concentration; China's 40% self-supply target; mutual-leverage argument
JSR acquisition by JIC; Japan's November 2025 export controls
Three qualified EUV resist suppliers; re-qualification as process change
12–36 month qualification timeline
3–5 year, tens-of-millions entry cost for new suppliers
2026 naphtha-driven photoresist supply warnings
Naphtha disruption; Chinese capacity additions
2023–2026 Chinese export controls; IEA magnet-block estimate
Price doubling; China-vs-Europe price divergence
Gallium/germanium share; exports to Japan at zero
China's rare earth mining vs. processing share
0.1% de minimis rare-earth licensing threshold
US-specific gallium/germanium/antimony ban suspension to Nov 27, 2026
Status of Chinese export-control suspensions
November 2026 deadlines; truce extension
China's EUV prototype (Dec 2025)
Status of China's EUV efforts
Japan's share of global photoresist (~75%)
Japan's share of EUV photoresist (>85%)
