Reverse engineering is not as simple as you make it sound. If you don’t know why the components work the way they do then reverse engineering doesn’t work, and even if you do know why you have to be able to make every part to the same specification and quality or your instrument is no good. And if you do know why they work the way they do then you’re frankly better off just assembling your own design rather than waste time trying to reproduce someone else’s. You’re trying to build something that can be used commercially at high frequency, so quality matters here. In China’s case they already developed all the component technologies before the ban. It’s simply a matter of assembling those technologies into a complete system and then figuring out the necessary optimizations, and the systems level optimization work is something that would be time consuming for a reverse engineered solution anyways. Honestly reverse engineering is never a good answer when you need to build a product. It also doesn’t give you a shortcut or instant results, and can end up being an even bigger time sink than studying first principles and then developing your own design from scratch. It’s something you do when you have no other option, and you need a starting sample to study the technology you don’t know. That’s not where China’s technological base is today.
Chinese semiconductor industry
- Thread starter Hendrik_2000
- Start date
- Status
Can't wait until the GOP calls for UN Inspectors to be let into China to make sure that they're not using any Western tools for manufacturing any part of the Huawei Mate 60, which is basically a WMD. West's Mass Destruction.
Last thing I remember on this topic was @hvpc being quite pessimistic. CXMT is like... 4-5 generations behind at the moment? And I think at least a generation and a half away from DDR5.
But do correct me if I'm mis-remembering. Either way, this does look like the biggest bottlenecks for China at the moment from a very layman POV.
SIOM and CIOMP's EUV-related patents for China-only, a market where ASML EUVs are not even permitted to be sold. ASML could care less.... it is only a concern when Chinese EUVs are exported abroad in overseas markets where it directly competes against ASML EUV, for example: TSMC, Intel, Micron, Samsung fabs in overseas.... China has no intentions of exporting EUVs abroad, and those fabs would never be allowed to install Chinese EUV in first place, so there is no IP conflict concerns. If for some reason ASML decides to challenge SIOM and CIOMP in court, then worse case, China can invalidate the patents on grounds of national security (i.e., compulsory licensing, infringement exemptions, etc...)
I think ASML would still care. Because when China makes its EUV and Chinese fabs expand and grab more market share, that will impact other fabs like TSMC and Samsung which are ASML customers. So in the end it will affect ASML bottom line too.
I used reverse engineering as an extreme example to inquire about China's capabilities in developing the sub-components of a LPP EUV source. But to give a concrete example for my IP scenario, when Chinese companies need to design droplet generators and laser droplet targeting subsystems, would existing IP by foreign IP for these tricky subcomponents make it meaningfully more difficult or delay China's progress?
I recall someone mentioning China's droplet generator is at a different frequency than that of AMSL's (50khz), could it be possible this design choice is partially influenced by IP rights?
Are you sure China has no intentions of exporting (LPP based) EUVs machines and subcomponents abroad? China is known for decimating high tech companies, ASML is already a national target for China. Even if Chinese equipment manufacturers only sell to Chinese fabs, you can't say SMIC will never leave the mainland.
Taking a deeper look at SSMB technology, it seems like the biggest benefit of SSMB proposal over other accelerator based light source proposals is that is it much more energy efficient from the storage ring. I think China can commercialize a LPP based EUV source earlier than the SSMB EUV. China's SSMB project (at xiongan?) is probably better fit towards beyond EUV lithography at around 6.7 nm. This also fits more inline with the phrase 弯道超车.
Last edited:
I think you are looking at this question the wrong way.
For EUV lithography (and the semiconductor industry tech stack as a whole), a better way of thinking about it imo is:
1. If there was a legitimate impetus for China to pursue a fully domestic semiconductor industry tech stack, how long would it take for the various components, subcomponents, consumables, materials, and final products need to be developed, tested and verified and delivered for commercial use?
2. When did a legitimate, proper impetus for china to pursue a fully domestic semiconductor tech stack actually occur and become realized by the industry at large?
"Having the technical capabilities" can still mean in some cases a multi year or multi-decade long wait for a product to reach fruition (not saying it will take that long in this specific case).
If you're specifically asking about whether IP considerations may be inhibiting or delaying certain aspects of the EUV lithography or semiconductor tech stack development by causing them to pursue other alternative development pathways, I think that's a bit of a specific question and would need to examine some of the IP and patents one by one while also considering whether those would be things that China would care about in this context.
But if your question is being asked from the perspective of "why is it taking so long to develop EUV lithography" -- then refer to the above questions no. 1 and 2, IMO.
Where would SMIC go and still be competitive? There isn’t any viable options that doesn’t span decades into the future.
I'm asking because I suspect taking into IP rights into the design of every single element of the EUV tech stack does pose a meaningful delay (e.g., 2 years). Perhaps there isn't a specific subcomponent that needs to be radically redesigned due to conflicting IP, but small changes here and there really add up.
From a high level perspective I'm wondering if for example SIOM, CIOMP and CAS could have put together a few reversed engineered sub-commercial quality EUV machines that were basically reverse engineered copies of ASML's EUV lithography machines starting from say 2019, after Meng Wanzhou was arrested. These machines would be for internal Chinese use or for sanctioned entities like Huawei as a stopgap measure, with an expected delivery of 2024.
Meanwhile SMEE or some other actual Chinese commercial entity would be another team actually developing a fully domestic EUV lithography tech stack with their 100% of their own IP. I'd expect this team to take longer say expected delivery in 2027 due to more polished designs and independent IP considerations.
The first copy, then reinvent model of innovation is tried and trued, in China and over the world. Would the two team approach I described above have any advantages over China's current plan?
TMSC has fabs outside Taiwan, I don't see why SMIC can't build fabs or buy fabs outside mainland China in the next 10 years.
This is a very specific hypothetical situation that we cannot really answer without knowing what IP or patents are in play, and there are so many permutations of “reverse engineer” that we can’t really account for them all.
However I think the actual time, money and engineering resources needed to develop, test and verify “reverse engineered” or “patent infringed” products like you describe would still end up taking an overall very similar amount of time and effort to a domestic product (even if such a “reverse engineered/patent infringed” product is viable or sensible to develop in the first place, also remembering they have never actually received delivery of an ASML EUVL product).
That said, my personal feeling is that I do not think your suggestion would have been much meaningfully faster and if anything may have resulted in dividing of efforts, resources and money to two parallel projects. Heck it is even possible that a reverse engineered approach could actually be slower than a domestic approach.
Regardless of the source of IP, to actually implement, test and produce a product to be viable for use needs engineering effort, money and organisational drive. There are no shortcuts in that regard, and those all need time.
Like, China will probably have EUVL that is viable for commercial use sometime in the second half of this decade. I don’t think it is that necessary to consider specific ways in which they might be able to get their a couple of years sooner especially if it requires so many hypothetical unknowns that may actually require substantial resources, such that if they had actually been implemented may actually set overall development back compared to their current track.
In the long run, a difference of a couple of years isn’t that important.
Last edited:
Hvpc has always been pessimistic. He was pessimistic since the start of sanctions and will be pessimistic all the way to the first domestic EUV and beyond. I don't think the man has ever shared an opinion that wasn't pessimistic.
- Status