What if Moore’s Law is different in China?

Ever since the US started to place an embargo on supplying high tech chips and semiconductor production equipment to China, I’ve wondered what the ultimate effect of that policy would be;  in particular, whether it might be self-defeating.  We had the first intimation of this last September when Huawei’s launched their Mate 60 Pro phone.  It included a new 7nm processor, that appeared to compare favourably with the smaller geometry processors Chinese companies no longer had access to.  Last week we saw a second example, with the unveiling of DeepSeek’s R1 AI application, which seems to achieve similar results to ChatGPT, but using a fraction of the power.  That came as a bitter blow to the nascent VC funded nuclear industry, which had been angling for a future of fusion and fission powered data centres.  It also got Wall Street panicking about whether it might have misjudged the value of AI.

To anyone who has read Michael Rosen’s children’s book, “We’re Going on a Bear Hunt”, this news shouldn’t have come as a surprise.

The premise of “We’re Going on a Bear Hunt” is how to cope with obstacles.  As each new on gets in the way, the family resort to the mantra of “We can’t go over it, we can’t go under it, we need to go around it”.  It’s a great book, which is enormous fun to read, but obviously aimed at a slightly older mental age than that of most US politicians. 

If you tell a bright engineer that they can’t do something, they tend to start thinking about how else they can achieve the desired effect.  When the US placed technology embargoes on China, it appeared that their only concern was that Chinese companies would find some way of smuggling the technology they needed across the border, or spend years trying to catch up.  Instead, what we are seeing is what should have been obvious – Chinese engineers would look to find alternative technologies.

There’s a very common, Panglossian approach of believing that whatever bleeding edge technology we’ve developed has to be the best way of doing something, and that development is generally a small, incremental process, without any Eureka moments on the way.  It’s exemplified by the Silicon Valley mantra of “Fail Fast, Fail Early, Fail Often”.  It’s a principle that often works, but implies a fairly blinkered approach to the way you’re going.  What we’re seeing with the Mate 60 Pro and DeepSeek is a different attitude of climbing out of the silo of accepted practice and asking how else something could be done.  In neither case does it appear to have been the most obvious approach, but in both cases it has moved China beyond where the US expected them to be.

What will be interesting is to see how long it takes China to bypass the lithographic approach of ASML – the Dutch company that enables the most advanced chips to be made, using their world-leading extreme UV process.  Although few people are probably aware of it, the US Government  has managed to exert pressure on the Dutch Government to ban ASML from selling their most advanced equipment to China, depriving Chinese semiconductor fabs of emulating the ultra-fine features that can be achieved in neighbouring Taiwanese fabs.  Once again, the belief is that it will take years for Chinese companies to catch up.

 The reason that these lithography machines are important is Moore’s Law – the rule of thumb assumption that it is possible to double the number of transistors you can put on a silicon wafer every two years.  This progress allows chips to perform more calculations, whether that’s for computing, games or AI.  The increase is enabled by making the features etched into silicon progressively smaller and smaller, letting you squeeze more into the same space.

Making these smaller has long been a fight against physics.  Back in the mid 1980s, when I was working on plasma processes in the semiconductor industry, everyone was concerned about making these smaller than 1 micrometre, as that seemed to be the limit using visible light.  The industry played with multiple approaches, out of which Extreme UV won, and the others were discarded.  Today, we’re almost a thousand times smaller, attaining  3 nanometre features in the most advanced fabs, and the industry pondering a future of 1 nanometre. 

Getting to this point has been the result of highly focused, incremental development, which has largely abandoned alternatives.  There will certainly be a massive amount of research going on within China to reproduce the technology which ASML is no longer allowed to sell them.  But there is probably just as much going on to see if any of the alternative lithography approaches can be brought out of obscurity and developed into a viable alternative.

Around the world, politicians seem to think that their country needs to be a leader in technology.  History shows that technology doesn’t really work that way.  For all of the rhetoric about leading in silicon, Taiwan still makes the best chips, using Dutch equipment.  Despite regular political interference, that joint leadership has been the status quo for the last twenty years.  No amount of money being poured into their local semiconductor industries by the US or European Governments has changed that. 

China will catch up.  Putting policies in place which encourage lateral thinking may well propel them to leadership.  It time for politicians to understand the dynamics of technical development and start to think about how best to use the technologies rather than trying to keep them for themselves.  Maybe we should start by getting them to read a few more children’s books before they go to bed each night, and teach them that it’s good to share each other’s toys.