Chinese semiconductor industry

cctang said:

He has good grounds for believing this. Look into what happened in the space industry - European vendors developed an entire family of ITAR-free products (including complete satellites) to serve the Chinese markets, after the Cox report in the US led to a complete ban on any sort of US commercial space technology being sold to China.

If they built ITAR-free satellites for the Chinese market, I think they can / will build de-Americanized EUVs for the Chinese market.

Click to expand...

My reservations with the line of thinking that @tphuang has expressed, is that I believe that the US will find it very easy to use political pressure to actively prevent the European nations to develop de-Americanized products to begin with, let alone sell them.

The real lever of power being applied which prevents nations from selling their wares, isn't that certain products have a certain percentage of US IP or US technology (that's just a formality), but rather those nation's geopolitical relationship (i.e.: geopolitical vulnerability) with the US.
The US can set the percentage of "US technology/IP" of XYZ product to any level they want -- including to 0%, and they can still apply geopolitical power and pressure to nations to prevent them from selling specific products if they feel giving China access to said products would be detrimental to US geostrategic interests.
The US has shown it is willing to make extra-territorial laws for transactions that do not involve the US in any form, so this really would not be a big step forwards at all.

So, I think a more likely outcome for semiconductor and semiconductor-related equipment and technologies will be more akin to the ban on selling military equipment to China -- a wide spanning arms embargo that the US and its dependents all voluntarily ascribe to and where even the idea of selling such equipment to China results in condemnation and dog-piling onto the group which does so.
I could easily imagine the US making a law whereby they are able to impose sanctions/threaten geopolitical relationships with any nation (including/especially its allies) that has any companies under their jurisdiction that sell China semiconductor industry related equipment -- if anything, this is just one of the logical outcomes of the current US trend, and this isn't even the most extreme or severe potential outcome of the US trajectory.

The only thing which may prevent such outcomes, is if the industries in Europe and other nations like Japan actively lobby their own governments... But I think US pressure will be impossible to resist against, so that the governments in question will have to be forced to choose between their entire relationship with the US (and the US security/geopolitical structure overall) versus selling China domestic/de-Americanized semiconductor products, and I suspect they will not choose the latter.


In the long term, the only path forwards for China is to domesticate the entire, full tech stack, and/or to have parts of the tech stack with nations that are not geopolitically aligned with the US or geopolitically vulnerable from US pressure... but of course we all knew that.

That said, I'm sure there may be a time period in the future whereby some European companies can develop de-Americanized products and be able to sell China those products before the US makes laws and carries out pressure to stop them from selling de-Americanized products... but it's also possible that the US government seeks to go ahead of the curve and puts in legislation and applies pressure before European companies can even finish developing de-Americanized products, to pre-empt any equipment from reaching China at all.

Blitzo said:

My reservations with the line of thinking that @tphuang has expressed, is that I believe that the US will find it very easy to use political pressure to actively prevent the European nations to develop de-Americanized products to begin with, let alone sell them.

The real lever of power being applied which prevents nations from selling their wares, isn't that certain products have a certain percentage of US IP or US technology (that's just a formality), but rather those nation's geopolitical relationship (i.e.: geopolitical vulnerability) with the US.
The US can set the percentage of "US technology/IP" of XYZ product to any level they want -- including to 0%, and they can still apply geopolitical power and pressure to nations to prevent them from selling specific products if they feel giving China access to said products would be detrimental to US geostrategic interests.
The US has shown it is willing to make extra-territorial laws for transactions that do not involve the US in any form, so this really would not be a big step forwards at all.

So, I think a more likely outcome for semiconductor and semiconductor-related equipment and technologies will be more akin to the ban on selling military equipment to China -- a wide spanning arms embargo that the US and its dependents all voluntarily ascribe to and where even the idea of selling such equipment to China results in condemnation and dog-piling onto the group which does so.
I could easily imagine the US making a law whereby they are able to impose sanctions/threaten geopolitical relationships with any nation (including/especially its allies) that has any companies under their jurisdiction that sell China semiconductor industry related equipment -- if anything, this is just one of the logical outcomes of the current US trend, and this isn't even the most extreme or severe potential outcome of the US trajectory.

The only thing which may prevent such outcomes, is if the industries in Europe and other nations like Japan actively lobby their own governments... But I think US pressure will be impossible to resist against, so that the governments in question will have to be forced to choose between their entire relationship with the US (and the US security/geopolitical structure overall) versus selling China domestic/de-Americanized semiconductor products, and I suspect they will not choose the latter.


In the long term, the only path forwards for China is to domesticate the entire, full tech stack, and/or to have parts of the tech stack with nations that are not geopolitically aligned with the US or geopolitically vulnerable from US pressure... but of course we all knew that.

That said, I'm sure there may be a time period in the future whereby some European companies can develop de-Americanized products and be able to sell China those products before the US makes laws and carries out pressure to stop them from selling de-Americanized products... but it's also possible that the US government seeks to go ahead of the curve and puts in legislation and applies pressure before European companies can even finish developing de-Americanized products, to pre-empt any equipment from reaching China at all.

Click to expand...

This!

Geopolitical pressure is a prerequisite to the efficacy of extra-territorial export controls.

If US has no power on a specific government, then extra-territorial laws are moot, see China for instance. Hundreds of Chinese suppliers sell to Huawei, but US knows that, apart form putting them on entity list, any other action is impossible, because those firms are protected by Chinese government and US has no influence on the Chinese government.

OTH in Europe, especially starting with Biden administration and with the leadership change that occurred in Germany after Merkel, we are having a huge regression in terms of sovereignty. To be clear, it is 70 years that Europe lost sovereignty, but US's leash has been looser or tighter according to the period. Today it is very tight and is getting tighter, due to a political class, especially in Brussels, that acts like a real US asset (Von der Leyen for instance, she aims to become next NATO chief after Stoltenberg) instead of representing European citizens interests.

US is moving to stop European high-tech companies to sell in China, with the help of European politicians. But due to the way Europe politics work, this will not be quick, so China has some time to fully localize high-tech. The best defense Europe has against US is to slow down the process, see ASML, it took one year for US to really close that loophole, and as of today is still not clear how much time it will takes to really stop ASML.

But this is not the end of the story, there are already very visible sings in Europe of the next line of defense, when the the export controls on high-tech will fail (we are almost there, like 2/3 years there). The next line will be import bans. US will rise huge protectionists barriers on high tech from China, and will coerce Europe to do the same, so to limit access and revenues to Chinese high-tech firms on very rich markets. This will be the next battle.

cctang said:

He has good grounds for believing this. Look into what happened in the space industry - European vendors developed an entire family of ITAR-free products (including complete satellites) to serve the Chinese markets, after the Cox report in the US led to a complete ban on any sort of US commercial space technology being sold to China.

If they built ITAR-free satellites for the Chinese market, I think they can / will build de-Americanized EUVs for the Chinese market.

Click to expand...

I also thought this was a possibility back when this started. But the current EU leadership is just too dumb and will follow whatever the US tells them to do. We are long past the time when Europe had some semblance of autonomous policy making.

Back when the ITAR sanctions against China in space tech happened the Europeans thought they could have been next. So major investments were done to secure access to space, build the Galileo satnav network, etc. Europe already had prior experience with the US blocking European satellite launches and giving priority in launch to US satellites in the 1970s. And in the 1990s the US had already had several major trade spats with the EU as well.

Europe is supposed to get the "carrot" of an Intel fab in Germany, and Japan is supposed to get the "carrot" of Rapidus with GAA technology licensing from IBM. But the truth is both the European and Japanese semiconductor industries are increasingly dominated by foreign (namely US) capital.

Toshiba's NAND facilities aka Kioxia were bought by US corporate vulture Bain Capital after seriously dubious shenanigans with the Toshiba acquisition of Westinghouse's nuclear business. Westinghouse was hiding debts after they acquired a construction business that was building four nuclear reactors in the US suspiciously right before Toshiba bought them. And US company Micron went on an acquisition spree and bought Elpida memory (JP), Inotera (TW), and the intellectual property of Qimonda (DE). The rug was pulled from under the feet of the Japanese semi industry and they are just too dumb to do anything about it. US company Texas Instruments even bought two Japanese semi fabs from Spansion, packed their most advanced machine tools, and moved them to the US.

As for the European semi industry, you had ARM being bought by Softbank, and then IPOed in the US. Their leading edge high-end chip design is made in the US. It is basically a US company at this point. ARM core designs like Neoverse V2 cannot be exported to China. Good luck also exporting core designs with SVE2 in them. Yet the UK does not allow Nexperia to buy an old fab just because Nexperia is owned by a Chinese company. Qimonda is gone.

european_guy said:

Thanks for this.



Actually I don't understand how they can impose rules on ASML, like it is a US-company. But of course US rules are always based on power-play (especially with their so called allies), more than on international law.

Click to expand...

european_guy said:

ASML CEO spoke about limitations but only on few big firms, this part is totally missing in the document. Limitations seem to apply to all Chinese firms....probably there is some behind-the-curtain negotiation ongoing.

LAM Research stated that new regulations do not affect their business. I'd guess that, after last year fiasco, now they don't want US firms to be kicked out completely from China market...but this horse has bolted already.

Click to expand...

ASML is not an American company so the BIS export restriction does not apply directly to them. The latest BIS amendment to semicondutor export control may eventually impact Dutch's own export regulations, but at least it won't happen immediately.

The BIS export regulations dictates what equpiment require an export license. For those under the entity list, BIS will act under the presumption of denial. Export license to advanced Chinese fabs will be denied. Export to mature fabs do not fall under the EAR scope, but some equipment within the export regulation to these mature fabs still require export license.

european_guy said:

The most worrisome news is about lithography machines. New rules state that DUVi is banned in China. This means to cripple not only advanced nodes, but also 28nm node. If US succeeds in banning ASML, this will be a "success" for US hawks....and a cold shower for everybody else.

Click to expand...

What I expect to happen is the latest tightening of scope of ArFi scanner will impact exporting of U.S. made parts to maintain/fix ASML's NXT1980 scanners. Parts, for instance, related to the reticle stage, alignment system, and some lens element are built by ASML's US factory, so these most likely may not be shipped to China.

european_guy said:

Now it is really time from SMEE to come out with a viable commercial solution!

Click to expand...

You just can't put a date to break through technical barriers. Breakthrough do not adhere to some schedule or target date decision makers (or sideline cheer leaders like us) come up with. You don't just set a target to find the cure for cancer by the end of 2023 and expect the R&D guy to stick to that target.

european_guy said:

China semiconductor and SME community did a wonderful job: today they have all covered for 28nm and soon for more advanced nodes too, but there isn't yet a commercial litho machine. After many years of stricter and stricter export controls, now chickens came to roost: China badly needs localization of lithography machine. This is the last hurdle in the long way to self-sufficiency, but is also the hardest to conquer....and is needed now.

These new export controls will hardly be effective in the last months of 2023, so if no new DUVi machine is delivered starting from 2024, China will feel the pain maybe from Q1/Q2 2024.

Click to expand...

ASML shipped a lot of litho tools to China in Q3. Based on what I heard, they will ship just as many, if not more, tool in Q4 as in Q3. Dutch government had agreed to let ASML continue to ship NXT2050 & NXT2100 until the end of 2023. You can expect ASML to ship lots of these advanced ArFi tools to Chinese fabs in Q4 to take advantage of the extended deadline provided by the Dutch government.

@tphuang had indicated correctly that by 2024 we will have "enough" advanced ArFi tools. The current bottleneck at these advanced nodes are not ArFi litho tools but American made process, metrology, and inspection equipment. Until we can find alternative to U.S. process & metrology equipment, the bottleneck will not be with litho tools. So, I agree with him that focus of SMEE should be to not rush things but to focus on building good foundation at the 28nm node (easier technical performance requirement) and focus on making the first domestic scanner reliable, robust, and HVM worthy. SMEE doesn't need to rush to push the performance capability envelop as it's not the bottleneck nor priority for Chinese fab's immediate success.

european_guy said:

One final note: SMEE had plenty of time to develop a commercial solution, indeed they had more than 10 years time! Unfortunately, due to complacency and a wrong read of the future, they wasted much of this time with money-grabbing static prototypes, and started to work for real only in the last few years...now all China semiconductor world is suffering for their shorth-sighted approach.

Click to expand...

No. SMEE leveraged their learning from the initial SSA600 ArF litho tool and came up with iline litho systems (SSB500) targeting the backend applications. (Note this is why iLine has a 'B' in it's name & KrF has a 'C'....it's alphabetical order matching the order in which they developed the different wavelengths scanner). They chose an application with a lower barrier to entry and they did good as they dominate the domestic backend litho market.

So, I disagree they were complacent. It's just that ArF, ArFi have higher technical challenge and barrier to over come...and harder to commercialize. Can't fault them for going after something that could make them some money instead of dicking around with something for a long time that will cause them keep bleeding money they don't have.

hvpc said:

ASML is not an American company so the BIS export restriction does not apply directly to them. The latest BIS amendment to semicondutor export control may eventually impact Dutch's own export regulations, but at least it won't happen immediately.

The BIS export regulations dictates what equpiment require an export license. For those under the entity list, BIS will act under the presumption of denial. Export license to advanced Chinese fabs will be denied. Export to mature fabs do not fall under the EAR scope, but some equipment within the export regulation to these mature fabs still require export license.

What I expect to happen is the latest tightening of scope of ArFi scanner will impact exporting of U.S. made parts to maintain/fix ASML's NXT1980 scanners. Parts, for instance, related to the reticle stage, alignment system, and some lens element are built by ASML's US factory, so these most likely may not be shipped to China.

You just can't put a date to break through technical barriers. Breakthrough do not adhere to some schedule or target date decision makers (or sideline cheer leaders like us) come up with. You don't just set a target to find the cure for cancer by the end of 2023 and expect the R&D guy to stick to that target.

ASML shipped a lot of litho tools to China in Q3. Based on what I heard, they will ship just as many, if not more, tool in Q4 as in Q3. Dutch government had agreed to let ASML continue to ship NXT2050 & NXT2100 until the end of 2023. You can expect ASML to ship lots of these advanced ArFi tools to Chinese fabs in Q4 to take advantage of the extended deadline provided by the Dutch government.

@tphuang had indicated correctly that by 2024 we will have "enough" advanced ArFi tools. The current bottleneck at these advanced nodes are not ArFi litho tools but American made process, metrology, and inspection equipment. Until we can find alternative to U.S. process & metrology equipment, the bottleneck will not be with litho tools. So, I agree with him that focus of SMEE should be to not rush things but to focus on building good foundation at the 28nm node (easier technical performance requirement) and focus on making the first domestic scanner reliable, robust, and HVM worthy. SMEE doesn't need to rush to push the performance capability envelop as it's not the bottleneck nor priority for Chinese fab's immediate success.

No. SMEE leveraged their learning from the initial SSA600 ArF litho tool and came up with iline litho systems (SSB500) targeting the backend applications. (Note this is why iLine has a 'B' in it's name & KrF has a 'C'....it's alphabetical order matching the order in which they developed the different wavelengths scanner). They chose an application with a lower barrier to entry and they did good as they dominate the domestic backend litho market.

So, I disagree they were complacent. It's just that ArF, ArFi have higher technical challenge and barrier to over come...and harder to commercialize. Can't fault them for going after something that could make them some money instead of dicking around with something for a long time that will cause them keep bleeding money they don't have.

Click to expand...

Interesting, how many NXT2050 & NXT2100 do you think China have now and by the end of 2023?

I had an impression that China had none of those machines

antiterror13 said:

Interesting, how many NXT2050 & NXT2100 do you think China have now and by the end of 2023?

I had an impression that China had none of those machines

Click to expand...

SMIC had at least one NXT2000 and one NXT2050 before last October’s US BIS sanction. That’s what they used to fab Kirin9000S on their N+2 (7nm) process node.

I heard ASML was rushing to ship more than 10 NXT2xxx systems to Chinese fabs before the September deadline imposed by the Dutch government. Since then, ASML received an extension to ship those advanced tools until end of this year. ASML’s Q3 financial result suggest they shipped lots of tools to China. Their Q4 forecast suggest it’ll be more of the same. I think more than 10 of those advanced systems is reasonable and very likely scenario.

hvpc said:

ASML is not an American company so the BIS export restriction does not apply directly to them. The latest BIS amendment to semicondutor export control may eventually impact Dutch's own export regulations, but at least it won't happen immediately.

The BIS export regulations dictates what equpiment require an export license. For those under the entity list, BIS will act under the presumption of denial. Export license to advanced Chinese fabs will be denied. Export to mature fabs do not fall under the EAR scope, but some equipment within the export regulation to these mature fabs still require export license.

What I expect to happen is the latest tightening of scope of ArFi scanner will impact exporting of U.S. made parts to maintain/fix ASML's NXT1980 scanners. Parts, for instance, related to the reticle stage, alignment system, and some lens element are built by ASML's US factory, so these most likely may not be shipped to China.

You just can't put a date to break through technical barriers. Breakthrough do not adhere to some schedule or target date decision makers (or sideline cheer leaders like us) come up with. You don't just set a target to find the cure for cancer by the end of 2023 and expect the R&D guy to stick to that target.

ASML shipped a lot of litho tools to China in Q3. Based on what I heard, they will ship just as many, if not more, tool in Q4 as in Q3. Dutch government had agreed to let ASML continue to ship NXT2050 & NXT2100 until the end of 2023. You can expect ASML to ship lots of these advanced ArFi tools to Chinese fabs in Q4 to take advantage of the extended deadline provided by the Dutch government.

@tphuang had indicated correctly that by 2024 we will have "enough" advanced ArFi tools. The current bottleneck at these advanced nodes are not ArFi litho tools but American made process, metrology, and inspection equipment. Until we can find alternative to U.S. process & metrology equipment, the bottleneck will not be with litho tools. So, I agree with him that focus of SMEE should be to not rush things but to focus on building good foundation at the 28nm node (easier technical performance requirement) and focus on making the first domestic scanner reliable, robust, and HVM worthy. SMEE doesn't need to rush to push the performance capability envelop as it's not the bottleneck nor priority for Chinese fab's immediate success.

Click to expand...

Can Nextin from South Korea replace KLA tools till Chinese equipment manufacturers come with workable solutions? I think they are already supplying YMTC, JHICC, CXMT and SMIC.

hvpc said:

No. SMEE leveraged their learning from the initial SSA600 ArF litho tool and came up with iline litho systems (SSB500) targeting the backend applications. (Note this is why iLine has a 'B' in it's name & KrF has a 'C'....it's alphabetical order matching the order in which they developed the different wavelengths scanner). They chose an application with a lower barrier to entry and they did good as they dominate the domestic backend litho market.

So, I disagree they were complacent. It's just that ArF, ArFi have higher technical challenge and barrier to over come...and harder to commercialize. Can't fault them for going after something that could make them some money instead of dicking around with something for a long time that will cause them keep bleeding money they don't have.

Click to expand...

hvpc said:

SMIC had at least one NXT2000 and one NXT2050 before last October’s US BIS sanction. That’s what they used to fab Kirin9000S on their N+2 (7nm) process node.

I heard ASML was rushing to ship more than 10 NXT2xxx systems to Chinese fabs before the September deadline imposed by the Dutch government. Since then, ASML received an extension to ship those advanced tools until end of this year. ASML’s Q3 financial result suggest they shipped lots of tools to China. Their Q4 forecast suggest it’ll be more of the same. I think more than 10 of those advanced systems is reasonable and very likely scenario.

Click to expand...

I don't think we need to know the exact number of ASML NXT2xxxx machines they will receive, but based on the much greater % of China market revenue for ASML in Q3 & probably part of Q2/Q4, they clearly delivered a lot of expensive ones. Enough for whatever expansion plans SMIC has for the next few years.

So we may have question of what to do about SMIC's mature process like 28/40nm, since they badly need to expand there too.

Based on what I heard before and please correct me if I'm wrong here, but it's probably easier for a new SMEE Arfi scanner to do trial production & achieve satisfactory yield on 40nm process vs 28nm. I assume they've also stocked up 1970 & 1965 at Jingcheng & Shenzhen fabs during this time.

So, I just don't see it makes sense to speculate what Dutch govt might do in 1 or 2 years. More important is what ASML actually can ship to China right now and to which fabs. And whether or not it can replace American parts.

As discussed with the Japan sanctions, they don't have anything in their law like what you see with what BIS controls. They've already granted licenses. Assuming that they will have same type of sanctions as US govt is just not reaonable

None of this is to say validation of domestic equipment should slow down, but it is okay to continue to use foreign equipments. And we should expect that they do that

I agree with everyone, I believe that China should not solely rely on the idea that European or Japanese supply chains will de-Americanize the US tech in the long-term because it's risky and unreliable solution. This process is largely beyond China's control. While China may occasionally influence such shifts, as many have pointed out, these shifts are vulnerable to arbitrary US political interference. China's only reliable solution is domestic full-stack tech, with sanction loopholes/bypass as only a temporary stop-gap measure until domestics arrive.

However, I'd like to introduce a subtle distinction in this discussion. The US, while influential, isn't all-powerful. Its sway over European and Japanese allies has boundaries and limitations. Every time the US exerts its geopolitical influence, it uses up valuable political goodwill. For instance, pushing a tech ban that impacts non-US technology (0% US tech, outside Wassenaar) to achieve vague goals (no clear objectives) while simultaneously affecting domestic and allied financial interests and without offering substantial compensations, costs the US significant political goodwill. Would the US really prioritize this over other pressing issues? To put it another way, the US will likely be strategic about its challenges, perhaps saving such a comprehensive tech ban for critical situations, like those involving Taiwan A.R., rather than engaging in ineffective power displays during peace-time that is clearly back-firing. In other words, US will carefully pick and choose it's battles, at the politically convenient time.

Alb said:

Can Nextin from South Korea replace KLA tools till Chinese equipment manufacturers come with workable solutions? I think they are already supplying YMTC, JHICC, CXMT and SMIC.

Click to expand...

My response may be biased because it’s mainly based on what I hear from KLA.

Nextin product offering is not comprehensive enough to fullly displace KLA. It’s also still at the stage where they are working with fabs to prove their equipment is good enough.

I personally think it’s reasonable to think Nextin may win some order here and there but they are still a bit far from challenging KLA’s dominance. Just my opinion.