Good find. By capacity in 200 mm equivalents we find a different picture - the capacity has been increasing.
They expect an expansion to 19% by 2024.
Chinese semiconductor industry
- Thread starter Hendrik_2000
- Start date
- Status
They expect an expansion to 19% by 2024.
I don't think it would be a good idea for China to initiate a move of pushing out ASML or Nikon. It would go against China's messaging of trying to work with international partners. Which in the long term is not good for their goal of selling technology to rest of the world. China does not benefit from protectionism.
But yes, I think 15 to 20 Arf/Arfi scanners is a nice volume to start off, but they will want to raise that up. China wants to be sufficient in mature nodes and it will require a lot of expansion in the next 5 to 10 years.
are you only getting total from adding up SMIC, Huahong and Nexchip? Does this include IDMs?
A lot of the revenues in China are split with local governments and many lower revenue power/industrial chips fabs that are now expanding rapidly. For example, Shanghai Jita, SMEC, Cansemi, CR Micro, Silan Micro, CRRC and BYD.
As for you other question, we already know the answer for SMIC is yes. It was quite transparent in saying that many of its American customers are hesitating in continuing to do business with them due to uncertainties from the sanctions. We know that American customers represent 20% of their revenue base. IIRC, Qualcomm is their #1 customer. I'm sure that will have major impact with their revenue in the next couple of quarters until they can either persuade the American customers to keep doing business or develop process for new domestic customers. As you can imagine, changing customers for your existing production line will lead to some issues. Losing access to certain American tools is only likely to show up in their future expansions. We don't know the affect. Although, I think it's reasonable to assume that they are one of the most likely to be affected by the US sanctions in terms of production.
Huahong does seem like the least affected by sanctions since they typically deal with mature chip production and serving domestic customers.
US would force both TSMC and Samsung to move a major portion of advance semiconductor plants no matter what. Whatever incentives the South Korean government came up with won't be enough to deter such move. US is literately stealing and force technology transfer both SK and Taiwan's crown jewel but both countries can't do anything.
In addition, China is investing a tremendous resources to cut reliance of mid to low end chips from other countries. In a few years, both TSMC and Samsung would be significantly weakened. Moreover, it seems many companies have lost interest in using more advance chips due to the expense and lack of urgency which would also severely close the gap and the advance of chips.
In addition, China is investing a tremendous resources to cut reliance of mid to low end chips from other countries. In a few years, both TSMC and Samsung would be significantly weakened. Moreover, it seems many companies have lost interest in using more advance chips due to the expense and lack of urgency which would also severely close the gap and the advance of chips.
The two list represent different things:
1. like you said, Trendforce is based on revenue / Knometa is physical wafer count
2. Trendforce list is only for foundries / Knometa list is for all sectors (foundries/IDM/memory)
3. Trendforce list is based on company’s HQ location / Knometa is based on fab location
Physical wafer output does not take into account the value of the wafer (e.g. not all wafers are the same; one 7nm wafer should be worth more than one 90nm wafer). So, I’m the industry, we do look at both numbers, but the revenue number reflects the overall value and hence better for us to extract business opportunities from OpEx spending, and is a better tarting point for projecting future business opportunities from the top-down.
Chinese foundries accounting for 10% WW market share at the moment. But the WW share of DRAM, 3D-NAND, and IDM would be a lot less.
There was an ICInsight analysis that showed revenue share of all fabs in China to be ~16%, but once removing Samsung/SK Hynix/Soidgm/tsmc/UMC then that figure drops to single digit, I think it was ~6%.
Anthis, there are many different ways to analyze data or to report data that fits our narrative. Looking at the Knometa list would make us feel we are closer to the MIC2025 target. But, a revenue ranking has been the more standard way to look at this thing. The motivation for MIC2025 is not because we import lots of chip, it’s because we spend more money importing semiconductor than oil. Motivation is to cut down on $$ semi import so we can boost the GDP. It’s all about the $$, number of wafer output is just a mean to get there.
The solution is very simple for the likes of Samsung of Korea.
They should choose China. Full stop. That is their biggest market and the market with the most potential. Nothing in this world can offer that to them.
If Samsung choose America, then two things should happen. One they lose their most important market China. Two is they might not reap the small benefits from America because TSMC already there, along with other US companies.
That's business. Make a mistake, and you can lose a lot.
The problem is the Koreans. They're stubborn. They should have been more flexible, and reacted quickly. But they are Koreans. They dither, assumed China cannot catch up to Korea (when Taiwan helped Samsung with tech), and they did not want to face an angry America.
We live with the choices we make.
this is the foundries ranking only. Per my response to @european_guy, we’d have a even smaller share of the IDM (memory and non-memory) space.
that’s not how things work. under standard international accounting rules, all revenue is reported usually by the controlling company. After removing all the expenses, the net profit is than divvy‘ed up between the controlling party and subsidiaries. So, for example revenue reported by SMIC would represent $ sale from all their wholly owned and JV fabs.
Fabs you listed play mostly in the IDM space and at the moment also represent a minor amount.
so far US has only “forced” tsmc to commit to roughly 10% of their N, N+1 (final) total capacity. I wouldn’t call this ‘major portion’.
mid-to-low end demand will increase over time, but these are lower margin lower profit business. I have a hard time understanding the logic behind how we can weaken tsmc Samsung.
Yes, we take away bigger piece of pie in the mid to low end, all the while tsmc and Samsung still take most the share of a even bigger piece of the pie. Revenue is important, it more importantly, the net protect matters, too. It’s hard to dent their business if we take command of small revenue and lower profit margin business.
Few months ago, I already point out the gap between us and the leaders will only widen as we don’t get to participate in the higher revenue/higher margin segments. Them, Oct 7th happened, and it made the situation worse. In this business, more CAPEX spend equals more profit. We sit here and admire about SMIC’s 340K wpm of 28nm and up that would cost more or less $30ishB. Do you realize tsmc will invest $12B for a single phase that’s 25K wpm of 5nm…they have plans for total of 8phases. Do you see the difference in magnitude?
I’d say, focus on ourselves building good foundation through expanding and getting foot hod mid to low end. But let’s not make outlandish claims or thinking we will be able to compete or hurt the industry leaders. tsmc will continue to be a cash cow in the foundries space, and Samsung will continue to dominate DRAM/3D NAND (as YMTC CXMT expansion plan is on hold, and they do not have enough wafer capacity or even more importantly bit capacity to impact Samsung much).
I think it remains to be seen how US semiconductor sanctions will impact YMTC and CXMT. I am concerned about CXMT lack of progress in DDR5 and higher density memory. But that has had little to do with the sanctions. Unlike them, YMTC has done fine with product development, the only question is if the US sanctions will impact their expansion. Like I said here some time ago, I think they capped their second fab recently, we will see if they will move equipment in and expand production or not. A lot of their equipment can be sourced in China anyways.
For the Memory portion of IDM’s, here’s the Q3 results:
As expected, these two segment is in the mist of the down cycle. But the Q/Q drop is quite drastic even for a “normal” cyclicality. even though the down turn in moemory was anticipated, the magnitude of the revenue hit is a bit surprising. Memories are commodities, the sooner it bottoms put and reach supply/demand parity, the quicker these segment would rebound. With that said, the inventory situation hasn’t improved as much in Q3 despite the revenue hit.
the industry still expect recovery in 2023. The question is how fast and how big of a rebound we will see this year.
Right. CXMT had trouble keeping up with the big boy even without the sanctions. We can blame it on it not being able to land EUV scanners. BUT, Micron was able to fab D1a without EUV, using exact systems CXMT has. This speaks to the difference in capability of the new kid on the block, CXMT, and seasoned veteran like MICRON that has a knack at squeezing every bit of capabilities out of a given equipment.
CXMT is one of the rumored fabs that had and is considering changing their business model. They may have to convert part of their capacity for foundry a la PSMC (foundry of logic and specialty memories), maybe even have to split up the current company/fabs to support the new model.
YMTC‘s expansion will be limited. Domestic equipments are not all adequate to support their original expansion.
But, if you/we really believe domestic equipments are as ready as we are hyping on SDF, then I guess there’s nothing for us to worry about on SDF. Unfortunately, we in the industry hear a different narrative from the fabs themselve. Well, we will see how things plays out this year. If domestic WFE suppliers could really quickly shape up to fill the gap then no problem, but if they can’t, then YMTC‘s future is also not very good. Not as good as the outlook prior to Oct 7th, but not as doom and gloom like that Bloomberg article someone shared earlier.
I’d give it a few months for domestic players to sort things out. We will know better one way or another before middle of this year.
- Status