The new numbers
In early October 2026, news outlets reported on a Washington think-tank study. It estimates that Chinese fabs had acquired 343 immersion deep ultraviolet (DUV) lithography machines by early 2026, about 270 of them ASML Twinscan NXT:1980i units. The report says China bought 90 of those machines in 2024 and 89 in 2025, at a cost of more than $13 billion (Seoul Economic Daily). The tools are older than EUV systems. They can still be pushed to 7nm-class logic and advanced memory for AI processors, according to coverage of the study (TrendForce).
The think tank is the Centre for Technology & Statecraft (CTS), a Washington-based group. These are estimates, not customs data, and should be read that way.
The report's conclusion is that Dutch licensing is the gap, and that Congress should close it with a China-wide ban.
The strongest case for the hawks
The case for tighter controls deserves a fair statement. The Netherlands has required national authorisation for advanced DUV lithography exports since September 1, 2023. On September 7, 2024 it widened that requirement to more equipment types. The government says it assesses each application case by case, so this is a licensing regime and not a ban (Dutch government). Minister Reinette Klever justified the measure by pointing to "increased security risks" and the potential military use of the resulting chips.
A case-by-case regime has a structural weakness. If the policy goal is to deny China leading-edge capability, every approved licence is a decision that can be gamed. Buyers can order ahead of expected tightening, which stockpiling in 2024 and 2025 would be consistent with. The 343-tool figure, if accurate, suggests licensing worked as a throttle and not as a wall. Given that 270 of the tools are the most capable NXT:1980i model, a security hawk can reasonably say the Dutch measure did not constrain what it was meant to constrain.
What the MATCH Act would do
The Multilateral Alignment of Technology Controls on Hardware (MATCH) Act was introduced in the House on April 2, 2026, with a Senate companion on April 8. It cleared the House Foreign Affairs Committee on April 22 (The Next Web). Its core mechanism is a 150-day deadline for the Netherlands, Japan and other suppliers to align their controls with US ones. If they do not, the US would extend its Foreign Direct Product Rule to allied tools built with American technology. The bill also reaches servicing of tools already installed in Chinese fabs (ChinaTalk).
The commercial stakes are large for ASML and for the Dutch economy. ASML has guided that China will fall to roughly 20% of its revenue in 2026, from 33% the year before. Enactment would push that figure lower still (The Next Web).
Why the ultimatum is the wrong instrument
The stockpile data supports reforming Dutch licensing. It does not show that a unilateral deadline is the best way to do that.
First, the diagnosis points to design, not sovereignty. The Netherlands has already tightened its rules twice and has a national process that updates through Government Gazette notices. The Dutch government's own export-control page shows a further modification scheduled for April 1, 2025 (Dutch government). A regime that moves this way can be made stricter through allied negotiation on defined technical criteria. It does not need a 150-day threat that treats an ally as a violator.
Second, the extraterritorial route carries costs the stockpile numbers do not measure. Extending US jurisdiction over allied products invites retaliation and pushes allies toward their own supply chains. ChinaTalk's analysis, otherwise sympathetic to the bill, concedes that proponents see it as political "cover" while opponents fear strained alliances. It also flags that the bill's servicing provisions are murky. Chinese engineers already employed by ASML and Tokyo Electron in China could continue maintenance under different arrangements, and it is unclear whether component suppliers such as Gigaphoton would be covered (ChinaTalk). A broad statute with unclear enforcement could impose heavy costs and deliver less control than advertised.
Third, a blanket China-wide ban is a blunt tool. Controls work best when they are narrow, evidence-based and tied to end users. Proportionate regulation would name the fabs and applications of concern, publish the criteria for denial, and keep licensing predictable enough for firms to plan around. The goal is to restrict capability that feeds military and frontier AI systems. It is not to sever an entire market on a signal that sales have been too high.
Fourth, the report's own recommendation is aggressive. As news coverage relays it, the authors urge a full DUV blockade beginning in 2027. The authors are former US export-control officials, and their analysis is a policy advocacy product. It should be weighed alongside Dutch licensing records, which the public cannot see. The Netherlands has not published how many licences it granted or refused, and that opacity is a legitimate target for reform.
A better path
The stockpiling evidence does support a few concrete reforms that a pro-innovation, pro-trade position can accept:
- Publish licensing statistics. Aggregate counts of approvals, denials and end-user categories would let allies and the public test whether the regime restrains what it should.
- Replace ad hoc case-by-case approval with presumptions of denial for named advanced-node fabs, with transparent exceptions for legacy-node commercial use.
- Coordinate with Japan and the US through a standing technical forum and not a statutory countdown, so that definitions of covered tools are shared and updated.
- Treat servicing and parts as a separate, narrowly drafted question, since that is where the MATCH Act is least clear.
These steps would address the gap the report identifies. They would avoid handing Washington a lever to override Dutch rulemaking, and they would keep the Dutch tech sector's legal certainty intact.
What to watch
The Dutch government decides how to respond, and the House has not yet voted on the MATCH Act. If Congress moves, the 150-day clock would give The Hague little time to rewrite its licensing rules. The Netherlands is better off acting first, publishing its data and tightening on its own criteria, than having them dictated.