Core Summary
Rand Corporation’s 2026 report compared two prediction methods (expert Delphi vs. crowdsourcing) on China’s ability to make advanced lithography machines—key tools for making chips. The questions: Can China mass-produce a DUV machine (140 wafers/hour) by Jan 2026? And an EUV machine (90 wafers/hour) by Jan 2030? Experts (6 people) were pessimistic (25% DUV,10% EUV), while the crowd (80 regular people) was more optimistic (37% DUV,45% EUV). By the report’s release, SMEE’s DUV machine was in testing but not fully ready—so experts were closer. However, the report had big gaps: it ignored China’s existing ASML DUV machines (which can make advanced chips via multi-patterning), used too few experts, and the crowd lacked deep tech knowledge.
1. Rand’s Experiment: Pitting Experts Against the Crowd
Rand tested two ways to predict:
- Delphi Method: 6 experts (AI researchers, China policy advisors, semiconductor supply chain pros) discussed anonymously in 3 rounds over a day to reach a consensus.
- Crowdsourcing: 80 regular people (half with high accuracy on Rand’s platform) had 5 weeks to guess and adjust their answers.
The questions had strict metrics: For DUV, it needed to produce 140 wafers per hour (a measure of speed) and be mass-producible. For EUV, 90 wafers/hour. The results? Experts were way more downbeat, especially on EUV—they thought it was a 10% shot, while the crowd thought almost half the time it would work.
2. Reality Check: Did China Hit the DUV Target?
By Jan 2026 (the DUV deadline), SMEE’s SSA800 DUV machine was in testing at SMIC (China’s top chip maker). Some sources said it could do 200 wafers/hour—beating Rand’s 140. But wait: Rand’s definition of “mass-producible” isn’t just speed. It’s about:
- Alignment precision: Can the machine stack chip layers perfectly (like putting a postage stamp on a moving target)?
- Yield: How many good chips come out (not just one or two, but enough for a factory)?
- Stability: Does it run 24/7 without breaking?
- Mass delivery: Can SMEE make hundreds of these machines?
ASML’s DUV machines are way better—300+ wafers/hour, super reliable, and proven in factories. So SMEE has progress, but it’s not there yet. That’s why experts’ 25% guess was closer to reality than the crowd’s 37%.
3. Why Experts and Crowd Disagreed So Much
The main gap is understanding system complexity:
- Experts know it’s a puzzle: A lithography machine isn’t one tech (like a light source). It’s 1000+ parts working together—光刻胶 (special glue), optical lenses (smooth as atoms), moving parts (no shake), software (controls everything to nanometer precision). All have to be ready at the same time. Think: Building a plane—you can’t fly with just an engine; you need wings, fuel, and a pilot.
- Crowd overestimates speed: Many crowd members thought China’s political will and money could rush progress. But some things can’t be rushed—like polishing a lens to atomic-level smoothness. You can’t throw cash to make that take 3 months instead of 3 years.
4. The Report’s Big Blind Spots
The report missed key things:
- Narrow question: It only asked if China can make its own machines, but China already has 100+ ASML DUV machines. With multi-patterning (using the same machine multiple times to etch smaller lines—like drawing a fine line with a thick pen by going over it), these machines can make 7nm or even 5nm chips. The report ignored this—like asking if a country can make cars but forgetting it already has 1000 imported cars.
- Too few experts: 6 people is a tiny sample. One expert initially thought EUV had a 75% chance, but others convinced them to lower it to 10%. Was this a good consensus or just group pressure? We don’t know.
- Crowd lacks expertise: Most crowd members were young (under 36) with no semiconductor experience. Their guesses were based on China’s progress in 5G or EVs—not actual knowledge of lithography.
- Prototype vs mass production: The crowd thought China’s EUV prototype meant fast progress, but a prototype is like a concept car—you can’t buy it in a store. It needs years of testing to be mass-produced.
5. The Real Strategic Question Rand Missed
The report focused on “can China make its own machines?” But the bigger question is: Can China use its existing ASML DUV machines to make enough advanced chips?
China has hundreds of ASML DUVs. With multi-patterning, these can make chips for phones, cars, and AI. Even if EUV takes longer to make, China can still stay in the chip game by optimizing these machines. This is the real strategic issue—Rand missed it entirely.
Final Takeaway
Rand’s report is useful because it tried to use data instead of opinions, but it’s not perfect. The key lesson: China’s chip strategy isn’t just about building its own machines. It’s about making the most of what it already has. And that’s the question we should be asking— not when China can make EUV, but how well it can use its existing tools to keep up.