Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/index.php?threads/techanalye-assesses-china-as-3-years-behind-tsmc.20865/page-5
        )

    [addOns] => Array
        (
            [DL6/MLTP] => 13
            [Hampel/TimeZoneDebug] => 1000070
            [SV/ChangePostDate] => 2010200
            [SemiWiki/Newsletter] => 1000010
            [SemiWiki/WPMenu] => 1000010
            [SemiWiki/XPressExtend] => 1000010
            [ThemeHouse/XLink] => 1000970
            [ThemeHouse/XPress] => 1010570
            [XF] => 2021770
            [XFI] => 1050270
        )

    [wordpress] => /var/www/html
)

TechanaLye assesses China as 3 years behind TSMC

That wasn't me.
My apologies. Talking to multiple people here.

Most of that cost to develop EUV was salaries in the US, Europe and Japan as well. The Chinese would be paying Chinese salaries. Let us say half the salary expense. Even if EUV was developed in the West with $20 billion USD they could probably develop it with $10 billion USD.
I would say that the cost is difficult to estimate, but likely much more than $20b. You do have a point in that Chinese salaries are lower. But EUV was developed due to multiple fortunate developments. Unfortunately for the Chinese, they would not benefit from these by developing from scratch. For example, if you win a scratcher (lotto) after buying several, you wouldn't only count the cost of the winning ticket. There were other companies which tried to develop EUV as well, but failed (like the Japanese). If China wants to guarantee success, you'd have to count the costs of those failed ventures as well. In short, China could spend $100b and still fail to develop EUV.
 
Last edited:
Their C919 airliner program has already cost them over $49 billion USD. So they can do it if they want to.
I should also mention that the C919 airliner program (and many other industries that China has succeeded in in the past) are different in that incremental gains are still rewarding. Like if the C919 fails, the effort they put into that still helps them with developing other airplanes. But in EUV, if they succeed in developing the light source but not the optics (or vice versa), then it doesn't really help them advance their chipmaking. Thus, the investment is wasted.
 
I would say that the cost is difficult to estimate, but likely much more than $20b. You do have a point in that Chinese salaries are lower. But EUV was developed due to multiple fortunate developments. Unfortunately for the Chinese, they would not benefit from these by developing from scratch. For example, if you win a scratcher (lotto) after buying several, you wouldn't only count the cost of the winning ticket. There were other companies which tried to develop EUV as well, but failed (like the Japanese). If China wants to guarantee success, you'd have to count the costs of those failed ventures as well. In short, China could spend $100b and still fail to develop EUV.
The Japanese lithography makers did not get the external funding from fabs like Intel and TSMC like ASML did. And their own chip industry was still stuck at 65nm for logic. Their NAND manufacturers need 28nm tops thanks to VNAND. Automotive requires about the same 28nm. So who would be buying these machines? Fujitsu to make a supercomputer every once in a blue moon? That is why it got cancelled, not any technical obstacles, I think they could have developed everything.

This is a news clip about a working Japanese 250W EUV light source for example. They have it supposedly.

Here is a presentation about their demo machines.
 
Last edited:
My apologies. Talking to multiple people here.


I would say that the cost is difficult to estimate, but likely much more than $20b. You do have a point in that Chinese salaries are lower. But EUV was developed due to multiple fortunate developments. Unfortunately for the Chinese, they would not benefit from these by developing from scratch. For example, if you win a scratcher (lotto) after buying several, you wouldn't only count the cost of the winning ticket. There were other companies which tried to develop EUV as well, but failed (like the Japanese). If China wants to guarantee success, you'd have to count the costs of those failed ventures as well. In short, China could spend $100b and still fail to develop EUV.
they could easily put some defense budget for this. you know how much money they spend on aircraft carrier.
this is actually what put them behind. too much spend on aircraft carrier instead of light sources. US has a solid program to support such things with darpa etc.
 
It’s straight off ASML financial statements. You can go look for yourself.
I'm not disputing that their capex is around $2 billion per year. But it doesn't mean what you think it means. Capex is the amount they spend on purchases of property, plant and equipment. How does that relate to how much it cost to develop EUV?

So who would be buying these machines? Fujitsu to make a supercomputer every once in a blue moon? That is why it got cancelled, not any technical obstacles, I think they could have developed everything.

This is a news clip about a working Japanese 250W EUV light source for example. They have it supposedly.

Here is a presentation about their demo machines.

Prior to EUV, Japan used to dominate the photolithography market (Canon and Nikon). Japan actually discovered EUV. But the despite extensive effort and collaboration with many Japanese companies, they failed to surmount all of the technical challenges. My takeaway from your links is that even if you are able to get to the point of working prototypes, it still does not guarantee that you can make something that is commercially viable.
 
I'm not disputing that their capex is around $2 billion per year. But it doesn't mean what you think it means. Capex is the amount they spend on purchases of property, plant and equipment. How does that relate to how much it cost to develop EUV?

I had another look at the financials. Call it $4b a year instead if you use the R&D budget. Or if you use the sum of Capex and R&D it's $6b. It could be $10b and it would still be small potatoes for China.

If you are saying the worlds largest economy measured by PPP cannot afford to develop EUV, it's silly. It's not a matter of cost at all, it's a matter of intellectual capital, but this is a country of 1.4b people so they will get there.
 
I had another look at the financials. Call it $4b a year instead if you use the R&D budget. Or if you use the sum of Capex and R&D it's $6b. It could be $10b and it would still be small potatoes for China.
See my other points above. The R&D was not all done by ASML. So you would have to add in the R&D from the companies they bought. Also, you would have to add in the R&D from the companies that they did not buy as well. (The ones that failed)
In addition, they also they also have suppliers that are still separate companies today. Add in their R&D costs as well.
 
See my other points above. The R&D was not all done by ASML. So you would have to add in the R&D from the companies they bought. Also, you would have to add in the R&D from the companies that they did not buy as well. (The ones that failed)
In addition, they also they also have suppliers that are still separate companies today. Add in their R&D costs as well.
Even with those, it's a small number. China's space program has outfunded its US counterpart in many areas, such as moon exploration and the space station. The whole initiative lies on Xi's mind, similar to how a CEO leads INTC. If you had Bob there, it would be much different from Grove, although both had sufficient budgets.
He might just want to play the underdog instead of crushing the US economy and employment, thereby bringing unnecessary troubles to himself.
 
Last edited:
Are you seriously implying that their EUV machines and all of the technology behind it was able to be developed by spending only a total of $2b per year in the past?
Successful EUV lithography probably required $100B in ecosystem development costs, between all the players. Remember, the fabs, the resist suppliers, the mask making suppliers, ASML’s sources like Cymer and Zeis, plus ancillary players like IP and EDA companies all had to invest substantially, in a coordinated way to get there. And that’s with highly profitable companies throughout the ecosystem pulling through - fabs still need to buy the EUV product once it is developed. SMIC, China’s hero foundry, is about a $7.5B / year company. TSMC is 10x that. Plus there is the time element - China could spend $100B on EUV and still be behind 5 years from now, requiring even more subsidies to catch up. And that doesn’t even fund the leading edge fabs.
 
Successful EUV lithography probably required $100B in ecosystem development costs, between all the players. Remember, the fabs, the resist suppliers, the mask making suppliers, ASML’s sources like Cymer and Zeis, plus ancillary players like IP and EDA companies all had to invest substantially, in a coordinated way to get there. And that’s with highly profitable companies throughout the ecosystem pulling through - fabs still need to buy the EUV product once it is developed. SMIC, China’s hero foundry, is about a $7.5B / year company. TSMC is 10x that. Plus there is the time element - China could spend $100B on EUV and still be behind 5 years from now, requiring even more subsidies to catch up. And that doesn’t even fund the leading edge fabs.
It's far cheaper to play catch-up. You don't have to try all the unknown options/paths.
Again, it seems Xi doesn't want to shake up his CAS bureaucracy and introduce some new blood, which could be the main issue.
 
Their C919 airliner program has already cost them over $49 billion USD. So they can do it if they want to.

While they deserve credit for assembling an airplane, it's important to acknowledge that all the vital parts are imported. Similarly, if China were to create an EUV machine using imported critical technology, could it truly be considered 'homemade'?
 
Here is an example of the typical hyperbole article written about China.

Even China's 1.4 billion population can't fill all its vacant homes, former official says
...
BEIJING, Sept 23 (Reuters) - Even China's population of 1.4 billion would not be enough to fill all the empty apartments littered across the country, a former official said on Saturday, in a rare public critique of the country's crisis-hit property market.
...
As of the end of August, the combined floor area of unsold homes stood at 648 million square metres (7 billion square feet), the latest data from the National Bureau of Statistics (NBS) show.
That would be equal to 7.2 million homes, according to Reuters calculations, based on the average home size of 90 square metres.
...
"How many vacant homes are there now? Each expert gives a very different number, with the most extreme believing the current number of vacant homes are enough for 3 billion people," said He Keng, 81, a former deputy head of the statistics bureau.

Just for reference, the average number of people per household in China is 2.76. So "enough for 3 billion people" would be 3 billion/2.76 = 1086 million vacant homes.

This is more houses than exist in China total, occupied and vacant. Double in fact.
The number of households in China is around 500 million.
And yet people eat up articles like this lunatic fantasy from Reuters.
I agree.

I'd be staggered if the average home size in China is as large as 90m^2 (these are almost all tower block apartments, aren't they ?). That's a big flat for 2.76 people in China ! You see smaller 3 bedroom houses than that in the UK.

So it's probably 10 million excess homes rather than 7.2 million.

And I guess those pictures we've seen of Chinese ghost cities and brand new blocks being demolished are all fake ...

But we're going way off topic here ...
 
It's far cheaper to play catch-up. You don't have to try all the unknown options/paths.
You’re just thinking about the R&D component for the EUV system. But you can’t just focus on that - you need the recreate the whole ecosystem that supports the full flow development, purchase and deployment of EUV. For instance, do you really get the full benefit out of EUV at 2nm if you don’t have GAA as well ? And how do you get even do EUV lithography unless you have EUV-tuned OPC/ILT to create the masks ? My view is that it is super expensive to build and sustain a completely parallel design and fab ecosystem for making chips using EUV. The sustain part is particularly hard and expensive, especially given the volumes SMIC is doing at the leading edge today.

That highlights the second, issue, the timing problem. If China really wanted EUV, it would need it today To sustain national champions like SMIC and Huawei. Instead, they are being forced to make allocation decisions between GenAI chips and smartphone chips due to inefficient and hard-wear usage of limited DUV equipment, and low yields. Apple will bank something like 30-40M chips before a new iPhone launch, to give them the head start they need to sell the usual 70M new model smartphones the first 5 months. Huawei is having problems banking even 2.5M chips / phones, though their volumes are lower (less than 10M units in first 5 months) but doing semi-yearly releases.

Apple's Chinese rival Huawei sees manufacturing snag ahead of flagship phone launch: report​

Huawei wanted to procure at least 2.5M chips ready before releasing Mate 70, and was planning for a September release. However, due to the current manufacturing capacity and productivity, the company will not meet that target, which is needed to meet expected consumer demand for the phone, the report noted.

SMIC uses outdated and inefficient tools to make advanced chips because of not being able to access essential U.S. equipment due to the ban. This has resulted in higher manufacturing costs and lower productivity. The Mate 70 chips SMIC did manage to make had a high defect rates. As many as 60% of the chips on each wafer were considered unusable, the report noted.

 
You’re just thinking about the R&D component for the EUV system. But you can’t just focus on that - you need the recreate the whole ecosystem that supports the full flow development, purchase and deployment of EUV. For instance, do you really get the full benefit out of EUV at 2nm if you don’t have GAA as well ? And how do you get even do EUV lithography unless you have EUV-tuned OPC/ILT to create the masks ? My view is that it is super expensive to build and sustain a completely parallel design and fab ecosystem for making chips using EUV. The sustain part is particularly hard and expensive, especially given the volumes SMIC is doing at the leading edge today.

That highlights the second, issue, the timing problem. If China really wanted EUV, it would need it today To sustain national champions like SMIC and Huawei. Instead, they are being forced to make allocation decisions between GenAI chips and smartphone chips due to inefficient and hard-wear usage of limited DUV equipment, and low yields. Apple will bank something like 30-40M chips before a new iPhone launch, to give them the head start they need to sell the usual 70M new model smartphones the first 5 months. Huawei is having problems banking even 2.5M chips / phones, though their volumes are lower (less than 10M units in first 5 months) but doing semi-yearly releases.

Apple's Chinese rival Huawei sees manufacturing snag ahead of flagship phone launch: report​

Huawei wanted to procure at least 2.5M chips ready before releasing Mate 70, and was planning for a September release. However, due to the current manufacturing capacity and productivity, the company will not meet that target, which is needed to meet expected consumer demand for the phone, the report noted.

SMIC uses outdated and inefficient tools to make advanced chips because of not being able to access essential U.S. equipment due to the ban. This has resulted in higher manufacturing costs and lower productivity. The Mate 70 chips SMIC did manage to make had a high defect rates. As many as 60% of the chips on each wafer were considered unusable, the report noted.


This reminds me of the debate 15 years ago on whether Chinese companies could produce quality cars without Western partners.

I think one way or another China will catch up or surpass western chipmakers. Either they will catch up on EUV, some other technology, or maybe in 10 years the focus will be on quantum, which China may be in the lead already.
 
This reminds me of the debate 15 years ago on whether Chinese companies could produce quality cars without Western partners.

I think one way or another China will catch up or surpass western chipmakers. Either they will catch up on EUV, some other technology, or maybe in 10 years the focus will be on quantum, which China may be in the lead already.
Seems like you are making stuff up - The US Big 3 may have questioned China's ability but there wasn't this kind of debate about China going it alone. Maybe you don't remember - China had plenty of enforced joint venture partners to teach them auto manufacturing and quality. Plus auto manufacturing plant and development tooling costs are an order or two in magnitude less than those for leading edge fab. I do give China a lot of credit for their success with EVs, but that was much more of a Blue Ocean venture - they decided long ago (late 1990s to 2000) to go where the rest of the market was not, and executed consistently to deliver. I think Elon's / Tesla's planting a factory in Shanghai in 2019 did help give the hundreds of EV companies at the time a more competitive target to shoot for, though, but that only helped drive world class products.

China may very well become the lead dogs in quantum because that is another Blue Ocean endeavor with a long time horizon. Who knows what that market will look like. But I don't think they are in a financial position to build and sustain a parallel semiconductor universe at the leading-edge.
 
This reminds me of the debate 15 years ago on whether Chinese companies could produce quality cars without Western partners.

I think one way or another China will catch up or surpass western chipmakers. Either they will catch up on EUV, some other technology, or maybe in 10 years the focus will be on quantum, which China may be in the lead already.
That's a poor example. The crucial point you missed is that China mandates foreign carmakers to establish a joint venture before accessing the Chinese market, leading to years of technological knowledge transfer to domestic auto producers.
 
You’re just thinking about the R&D component for the EUV system. But you can’t just focus on that - you need the recreate the whole ecosystem that supports the full flow development, purchase and deployment of EUV. For instance, do you really get the full benefit out of EUV at 2nm if you don’t have GAA as well ? And how do you get even do EUV lithography unless you have EUV-tuned OPC/ILT to create the masks ? My view is that it is super expensive to build and sustain a completely parallel design and fab ecosystem for making chips using EUV. The sustain part is particularly hard and expensive, especially given the volumes SMIC is doing at the leading edge today.

That highlights the second, issue, the timing problem. If China really wanted EUV, it would need it today To sustain national champions like SMIC and Huawei. Instead, they are being forced to make allocation decisions between GenAI chips and smartphone chips due to inefficient and hard-wear usage of limited DUV equipment, and low yields. Apple will bank something like 30-40M chips before a new iPhone launch, to give them the head start they need to sell the usual 70M new model smartphones the first 5 months. Huawei is having problems banking even 2.5M chips / phones, though their volumes are lower (less than 10M units in first 5 months) but doing semi-yearly releases.

Apple's Chinese rival Huawei sees manufacturing snag ahead of flagship phone launch: report​

Huawei wanted to procure at least 2.5M chips ready before releasing Mate 70, and was planning for a September release. However, due to the current manufacturing capacity and productivity, the company will not meet that target, which is needed to meet expected consumer demand for the phone, the report noted.

SMIC uses outdated and inefficient tools to make advanced chips because of not being able to access essential U.S. equipment due to the ban. This has resulted in higher manufacturing costs and lower productivity. The Mate 70 chips SMIC did manage to make had a high defect rates. As many as 60% of the chips on each wafer were considered unusable, the report noted.

World's largest contract chipmaker Taiwan Semiconductor Manufacturing Co. has agreed to invest 1.11 billion euros (US$1.4 billion) in chip equipment maker ASML in return for a 5 percent stake, and to bankroll research and development of future technology.

You cannot exaggerate the situation and promote your view. A more accurate assessment is TSMC's investment. $14 billion would be a close estimate for ASML's investment. Other parts of the ecosystem are cheaper. Let's say $20 billion in total, roughly the amount for two fabs.

SemiWiki could consider paying for more professional replies/analyses instead of people just leaving broken, biased comments.
 
Last edited:
While they deserve credit for assembling an airplane, it's important to acknowledge that all the vital parts are imported. Similarly, if China were to create an EUV machine using imported critical technology, could it truly be considered 'homemade'?
The Chinese also used to build high speed rail trains with imported components. Not anymore.
The Chinese have a roadmap to make their commercial aircraft with their own components. Should happen until the end of this decade. For example they are already testing their replacement engine the CJ-1000A.

As for lithography they are clearly behind but they seem to have multiple concurrent programs to develop basically everything from immersion DUV to high-NA EUV. The question is how fast they can do it.
For example they have research not only on LPP light sources for EUV, but also FEL and synchrotron light sources (SSMB).
 
Back
Top