China vs USA – Chinese Whispers

Somewhere the message became garbled.

• The US is concerned that China has obtained one of ASML’s EUV machines, but I suspect that what China has is a few subsystems that they have put together, which does not get it any closer to a working EUV machine.
• EUV is crucial because it is the key to making high-yield semiconductors below 7nm, and the US has unilateral control over who can access it.
• Hence, keeping China from accessing EUV technology is the foundation of its strategy to slow China’s rise as a technological and geopolitical power.
• However, this does not stop the rumour machine, and the US State Department appears to have heard that China has been able to get hold of one of ASML’s EUV machines, concerns which it has expressed to ASML.  
• I think that this is extremely unlikely, as these machines are massive and cost more than $400m each, and ASML knows exactly where all of them are and who is using them.
• This is because all of these machines are connected to ASML’s systems so that they can be monitored and adjusted in real time if necessary.
• A rumour that I do believe is that China has managed to cobble together an EUV machine and that it has been able to make some chips that actually work.
• To do this, it would have had to procure the subsystems that it is unable to make and then assemble them and write the software to make the whole thing work.
• It is this rumour that I think has been floating around the industry and has become distorted to make some think that China has procured a whole working machine from ASML.
• However, simply being able to assemble a machine that can produce a working chip is very far from being enough.
• ASML reached this stage somewhere around 2001, but it then took it another 10-12 years to get the system to work reliably enough to produce high volumes at economic yields.
• Hence, the fact that China may have a machine that works is a very far cry from China being about to produce millions of leading-edge chips to put into AI data centres.
• The net result is that I see no reason to change my long-standing view that China will not be able to manufacture leading-edge semiconductors in China for many years to come.
• It is this disadvantage that makes Chinese AI semiconductors so uncompetitive with their Western competitors.
• This means that rolling out AI services to 1.4bn users using domestic silicon is going to be a real problem and could result in China falling behind unless it can get compute from elsewhere.
• It is through importing compute from Western-powered data centres overseas that I expect China to plug the gap, but this is not without risk.
• The US is well aware of this workaround, and I suspect that when the Department of Commerce updates its rules in Q3 or Q4 of 2026, it will target this practice.
• This will cause real consternation and further constriction of the compute capacity in China, meaning that China will fall further behind.
• Its models are still likely to see a lot of use outside of the USA, but China’s ability to subsidise the compute to drive usage is already limited and may become much more so.
• This is why I continue to think that the West and the USA in particular have the advantage in the AI race that is currently being run.