Most people from the sidelines could see this being inevitable. I thought their best move could've been to act early, and sell their entire business between Saudi's and Russia. The latter has a big market to supply necessary silicon to, and the former has the capital and interest to hedge their oil wealth into. With new Russian Engineers adding to the company, Saudi money to fund Research, I actually figured they could've somewhat caught up to Intel or Samsung.
As it is, their fate is almost sealed. SMIC will miraculously produce chips that are identical to TSMC in the near future. And they will flood the industry with good alternative at a much cheaper price. Samsung will survive the flood with their own technology and connections. Intel will be an in-house thing. And TSMC will continue to get more profitable. But it is GloFo that will be facing the bankruptcy sooner than latter. Even if the authorities allowed them to get acquired by China's State, it's not like they need them or have a particular use for them, so even the Chinese don't want them now. Not for a fair price anyway.
"SMIC will miraculously produce chips that are identical to TSMC in the near future." Only if the Chinese -not necessarily SMIC- developed tech that could compete with ASML's cutting edge EUV scanners. And that is easier said than done.. Even if they cracked everything else (power source, masks, pellicles, EUV source etc) I doubt they could master the optics of Zeiss. They are truly one of a kind.
To print tighter and tighter nodes with DUV scanners they'd need to go beyond quad patterning to 5-, 6- or even 8-patterning. That is not practical or possible.
They will lose even more customers in the future. Let's not forget that China is building a small army of factories for mature manufacturing nodes, meaning many GF customers who don't need something better than 10nm could probably leave GF and go to a cheaper Chinese factory for their chips.
To reach 7nm would have cost GF far more money than it could afford. Given the small size of GF vs TSMC the investment would have been unlikely to ever be profitable. GF had no rational choice but to abandon sub 10nm. In the long term GF may well fail - however if it had tried to keep up with sub 10nm then it would have failed much sooner. Sub 10nm manufacturing is only for the big firms - with a single sub 10nm FAB costing well over $10billion a firm with profits of under $2billion/year can not afford one.
I looked up what GF is doing, and it seems that it does at least improves its processes (announcing 22FDX+ and 9SW RFSOI last year) and offers new abilities (lower power, smaller features, larger temperature range, ...). So it's not resting on its laurels. Still, some of the market is moving elsewhere, and it will be interesting to see how GF tries to handle this.
$1B profit on $7B revenue, not bad. They know their place in the market, and it can work for them. They can pick up depreciated assets from other fabs and move forward. With lower capital and R&D expenses they should be able to keep costs low and attract customers. It was a reasonable strategy to not go head-to-head with TSMC and IFS. Their biggest problem might be China attacking the low end of the market.
The Chinese (SMIC + others?) are not going to release their technology for foundry customers anytime in the near future. To do that would give western clients access to sensitive documents and technology information that is needed to do design work. That would quickly reveal that the Chinese are in violation of many patents and that they are stealing IP from other foundries.
What China will do, and that they are doing, is steal enough technology from the rest of the world to get cutting edge nodes working, start manufacturing their own chips with their own designs, and stop being customers of western companies. In automotive this is BYD, and pretty soon all the electronics in BYD cars will be manuf. in China as well. Then they will take their stolen IP and try to sell it to us in the west by undercutting prices.
I wonder would it be cheaper to build a 7nm fab now than a few years ago when Global Foundries abandoned it. You would imagine there would be more expertise available in the labour market and technological advances that could control costs. It might be worth another crack.
They should buy used equipment, they can't keep doing/milking this. I've seen not one article warning about China flooding the world with chips on older process nodes. I believe it and I think it will not be different on how they made solar PVs a lot cheaper. I've been suggesting Tesla/Elon that if they think AI is the future, then they should be able to fab their own chips, buying GF for example.
Elon is one revolting individual with far too much inflammatory commentary used in the vain hope of keeping the spotlight on himself so his flamboyance acts as free/low cost advertising to the small minds unable to comprehend his methods.
In a recent interview at Tech Tech Potato GF's CEO Explained to Ian that GF sold off one EUV machine and returned the other to ASML for some reduced amount. And Maybe OpenAI can look at investing there at getting an EUV machine and get GF some funding to produce any in-house OpenAI accelerator designs fabricated on some sub 10nm node. That way GF could continue to produce revenues and try and remain revenue positive and with some outside investment to get back in the leading edge node race if only to meet OpenAI's needed but maybe others as well!
GF will pick up and use 193 double patterning machines for 7 and as the used fabrication equipment becomes available on mature techniques GF will offer sub 10 nm. Customer audits are easy and foundries are clubs so as the need arises the outcome is obvious, migration. mb
This company was dead on arrival. They stopped investing hoping that clients stay with them forever. Good luck. Amd they even placed shares on the market, a real scam
They would get people with experience in the foundry business. A foundry needs to be able to manufacture semiconductors, but it also needs to be able to work with customers (and develop credibility with customers), which is something Intel didn't have to do when it was just manufacturing its own designs.
On the other hand, Intel already has partnerships with Tower Semiconductor and UMC, which is another way to address this issue.
Why would they want the dated process nodes with slim margin?
simple: there's a few diaper loads of chips, largely embedded, that would be microscopic on single-digit fabrication. you can look it up.
"Demand growth will be highest for logic chips fabricated at 20nm to 45nm nodes to meet the growing computing demands of centralized electrical/electronic architectures, while demand pressure at mature nodes larger than 55nm may ease. "
Yes, foundries need to offer a range of processes to benefit a diversity of needs. IFS+GF would be great from that perspective. You can bet China would throw up regulatory hurdles, but then IFS can just pull out of China.
The problem is… soon there will be Only one or maybe two manufacturers to make chips.. aka monopoli and we know what that means! There just is not companies that can keep up with the top dogs and soon there Only in top dog that can increase prices by 200%…
effectively, that time has come. if all user-facing companies have only a single source (or near enough to make no difference) for bottle-neck tools...
"ASML dominates the lithography market with an 82.9% market share followed by Canon and Nikon. Its market share is more than 8x larger than second-place Canon. " https://seekingalpha.com/article/4642779-asml-stoc...
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
26 Comments
Back to Article
Soulkeeper - Wednesday, February 14, 2024 - link
I guess "good enough" wasn't a good motto afterall.Kangal - Friday, February 16, 2024 - link
It's funny because it's true.Most people from the sidelines could see this being inevitable. I thought their best move could've been to act early, and sell their entire business between Saudi's and Russia. The latter has a big market to supply necessary silicon to, and the former has the capital and interest to hedge their oil wealth into. With new Russian Engineers adding to the company, Saudi money to fund Research, I actually figured they could've somewhat caught up to Intel or Samsung.
As it is, their fate is almost sealed. SMIC will miraculously produce chips that are identical to TSMC in the near future. And they will flood the industry with good alternative at a much cheaper price. Samsung will survive the flood with their own technology and connections. Intel will be an in-house thing. And TSMC will continue to get more profitable. But it is GloFo that will be facing the bankruptcy sooner than latter. Even if the authorities allowed them to get acquired by China's State, it's not like they need them or have a particular use for them, so even the Chinese don't want them now. Not for a fair price anyway.
Tams80 - Saturday, February 17, 2024 - link
Or they could just continue specialising as they have done.There are other chip makers who already do well enough doing so.
And if anything, being not based in an authoritarian state will allow them access to mkre contacts from democracies.
Santoval - Thursday, February 22, 2024 - link
"SMIC will miraculously produce chips that are identical to TSMC in the near future."Only if the Chinese -not necessarily SMIC- developed tech that could compete with ASML's cutting edge EUV scanners. And that is easier said than done.. Even if they cracked everything else (power source, masks, pellicles, EUV source etc) I doubt they could master the optics of Zeiss. They are truly one of a kind.
To print tighter and tighter nodes with DUV scanners they'd need to go beyond quad patterning to 5-, 6- or even 8-patterning. That is not practical or possible.
yannigr2 - Thursday, February 15, 2024 - link
They will lose even more customers in the future. Let's not forget that China is building a small army of factories for mature manufacturing nodes, meaning many GF customers who don't need something better than 10nm could probably leave GF and go to a cheaper Chinese factory for their chips.Duncan Macdonald - Thursday, February 15, 2024 - link
To reach 7nm would have cost GF far more money than it could afford. Given the small size of GF vs TSMC the investment would have been unlikely to ever be profitable. GF had no rational choice but to abandon sub 10nm. In the long term GF may well fail - however if it had tried to keep up with sub 10nm then it would have failed much sooner.Sub 10nm manufacturing is only for the big firms - with a single sub 10nm FAB costing well over $10billion a firm with profits of under $2billion/year can not afford one.
sutamatamasu - Thursday, February 15, 2024 - link
6nm is newer "12nm" for now, GF should searching new investors if want to survive.ET - Thursday, February 15, 2024 - link
I looked up what GF is doing, and it seems that it does at least improves its processes (announcing 22FDX+ and 9SW RFSOI last year) and offers new abilities (lower power, smaller features, larger temperature range, ...). So it's not resting on its laurels. Still, some of the market is moving elsewhere, and it will be interesting to see how GF tries to handle this.flgt - Friday, February 16, 2024 - link
$1B profit on $7B revenue, not bad. They know their place in the market, and it can work for them. They can pick up depreciated assets from other fabs and move forward. With lower capital and R&D expenses they should be able to keep costs low and attract customers. It was a reasonable strategy to not go head-to-head with TSMC and IFS. Their biggest problem might be China attacking the low end of the market.do_not_arrest - Wednesday, March 6, 2024 - link
The Chinese (SMIC + others?) are not going to release their technology for foundry customers anytime in the near future. To do that would give western clients access to sensitive documents and technology information that is needed to do design work. That would quickly reveal that the Chinese are in violation of many patents and that they are stealing IP from other foundries.What China will do, and that they are doing, is steal enough technology from the rest of the world to get cutting edge nodes working, start manufacturing their own chips with their own designs, and stop being customers of western companies. In automotive this is BYD, and pretty soon all the electronics in BYD cars will be manuf. in China as well. Then they will take their stolen IP and try to sell it to us in the west by undercutting prices.
Tilmitt - Thursday, February 15, 2024 - link
I wonder would it be cheaper to build a 7nm fab now than a few years ago when Global Foundries abandoned it. You would imagine there would be more expertise available in the labour market and technological advances that could control costs. It might be worth another crack.zodiacfml - Thursday, February 15, 2024 - link
They should buy used equipment, they can't keep doing/milking this. I've seen not one article warning about China flooding the world with chips on older process nodes. I believe it and I think it will not be different on how they made solar PVs a lot cheaper. I've been suggesting Tesla/Elon that if they think AI is the future, then they should be able to fab their own chips, buying GF for example.PeachNCream - Thursday, February 15, 2024 - link
Elon is one revolting individual with far too much inflammatory commentary used in the vain hope of keeping the spotlight on himself so his flamboyance acts as free/low cost advertising to the small minds unable to comprehend his methods.lucaB75 - Thursday, February 15, 2024 - link
Totally true. He is disgusting personPixyMisa - Thursday, February 15, 2024 - link
Blucher.<horses whinny>
FWhitTrampoline - Thursday, February 15, 2024 - link
In a recent interview at Tech Tech Potato GF's CEO Explained to Ian that GF sold off one EUV machine and returned the other to ASML for some reduced amount. And Maybe OpenAI can look at investing there at getting an EUV machine and get GF some funding to produce any in-house OpenAI accelerator designs fabricated on some sub 10nm node. That way GF could continue to produce revenues and try and remain revenue positive and with some outside investment to get back in the leading edge node race if only to meet OpenAI's needed but maybe others as well!Bruzzone - Thursday, February 15, 2024 - link
GF will pick up and use 193 double patterning machines for 7 and as the used fabrication equipment becomes available on mature techniques GF will offer sub 10 nm. Customer audits are easy and foundries are clubs so as the need arises the outcome is obvious, migration. mblucaB75 - Thursday, February 15, 2024 - link
This company was dead on arrival. They stopped investing hoping that clients stay with them forever. Good luck. Amd they even placed shares on the market, a real scamsmalM - Friday, February 16, 2024 - link
Last year TSMC's revenue from 7nm was $13 billion on fully amortized equipment. There is no way for GloFo to compete in 7nm.onewingedangel - Friday, February 16, 2024 - link
Wonder if IFS may make another offer at some point, they want customers and specialised processes and can provide leading edge.dotjaz - Friday, February 16, 2024 - link
Why would they want the dated process nodes with slim margin? GloFo has <30% gross margin. What could Intel possibly get from it?KAlmquist - Friday, February 16, 2024 - link
They would get people with experience in the foundry business. A foundry needs to be able to manufacture semiconductors, but it also needs to be able to work with customers (and develop credibility with customers), which is something Intel didn't have to do when it was just manufacturing its own designs.On the other hand, Intel already has partnerships with Tower Semiconductor and UMC, which is another way to address this issue.
FunBunny2 - Saturday, February 17, 2024 - link
Why would they want the dated process nodes with slim margin?simple: there's a few diaper loads of chips, largely embedded, that would be microscopic on single-digit fabrication. you can look it up.
"Demand growth will be highest for logic chips fabricated at 20nm to 45nm nodes to meet the growing computing demands of centralized electrical/electronic architectures, while demand pressure at mature nodes larger than 55nm may ease. "
https://en.flykingtech.com/$%7Bcityname%7D/article/detail/221.html
Blastdoor - Sunday, February 18, 2024 - link
Yes, foundries need to offer a range of processes to benefit a diversity of needs. IFS+GF would be great from that perspective. You can bet China would throw up regulatory hurdles, but then IFS can just pull out of China.haukionkannel - Saturday, February 17, 2024 - link
The problem is… soon there will be Only one or maybe two manufacturers to make chips.. aka monopoli and we know what that means!There just is not companies that can keep up with the top dogs and soon there Only in top dog that can increase prices by 200%…
FunBunny2 - Sunday, February 18, 2024 - link
effectively, that time has come. if all user-facing companies have only a single source (or near enough to make no difference) for bottle-neck tools..."ASML dominates the lithography market with an 82.9% market share followed by Canon and Nikon. Its market share is more than 8x larger than second-place Canon. "
https://seekingalpha.com/article/4642779-asml-stoc...