And it should be obvious from peevee's answer it's the latter. 200nm wafers? Nobody's that dumb, if peevee needed to spell that last bit out for you, you shouldn't be on AnandTech.
What many do not understood is that 7nm is a very expensive node both for R&D and for production point of view. GF simply can't sustain that expenses and for such is now relegated as foundry playing in B league.
For me in 5 years GF will be dead or sold to a bigger foundry that may be interested in their buildings, not their technology or it will play in very niche market (RF, low power cheap IC, anything that is CHEAP) where it won't be able to come out anymore to fab for high performance industry unless a complete change in materials and cheaper machinery will come.
In few years will be left with only 3 fabs capable of using latest PPs: TSMC, Intel and Samsung, and I'm not sure there will be a place in the market for all three of them.
Agree and I do see a place for all three fabs going forward. I wonder if they will start sharing IP once we see 4-5nm as its going to even more expensive than 7nm.
That was my thought, too, that we're likely to see more extensive sharing of IP between companies as we shrink processes further. That's not the worst thing in the world, especially if it helps keep all 3 in the market. Every process shrink from 180nm down to 7nm has resulted in at least one foundry dropping out of the "bleeding edge" market (with the exception of the 22-14nm shrink), and it's bad enough there's only 2 choices for a fabless company (like AMD, Nvidia, etc.) to place orders for the latest process.
I do think that lithography is coming to a point of diminished returns, so 12nm and above might work well for SSD, I/O, wireless, NIC, memory, etc...
Eventually they'll have to go to the next step otherwise they're done for. Everyone will settle in at 7nm, 5nm or 3nm, but we're still at the end of process improvements moving the industry forward.
There is a ton of money in RF, ADC/DAC’s,power management. There is probably 10X more revenue in those areas for the products I work on than the processor, so no reason they won’t be around awhile. The key is just to know what you want to be and focus on it. Seems like that’s what they’re doing.
10x revenue but 1/10 net profits. It is not by chance that TSMC started ramping up their PP investments only after ARM SoC using latest available PP boom happened. If you want to make real money you have to produce with the latest PP. Otherwise you are a general foundry like any other (and you may know that all foundris, TSMC, Intel and Samsung operate in the same RF and network, which is another big market, as well).
Declaring their debacle on the new PP, GF simply puts itself in the second lines of foundries on the global market and with diminishing profits (while being a quite large company) it probably will further go behind to 3rd ot 4th line.
In 5 years they be none you will know bout as a consumer and will probably be swallowed (or will merge with) other small foundries to try to create a better marketing scale value, that is essentially what makes foundries make money.
They stopped development of 7nm and beyond because costs are prohibitive and they could not compete. Their execution had been lacking for the past few years, especially in process. They had to *license* Samsung's 14nm process, instead of using their own, as planned. Process is only getting more expensive and competition is moving faster. They retreated on their own, differentiated processes. What nobody else offers. This is rather smart. They have a market to serve and the competence to do so, with less competition. Their future, 5~10 years, is assured. Beyond that, they may finally push for smaller sizes at a much lower cost (and similarly add their own twist to it, like on their 12nm processes). There is a place for such niche manufacturing, even if it is less glamorous.
"We since have confirmed that the graphics is indeed fused off. Intel’s official line is that this processor was released with a specific target market in mind, and it fulfils the role required. What exactly this market is, and at what price point, is still a mystery, even in 2019. However some analysts believe the graphics was a dodo out of the door due to uneconomically viable yields, as well as this chip not making any sense commercially for the product segment it ended up in – it was put into the market just to fulfil a promise to investors.
Moving from May to August, and no new processors or devices relating to Intel’s 10nm were announced. However Charlie over at SemiAccurate published knowledge that there were issues with Intel’s 10nm process as currently presented. He reports that Intel’s yields were sub-10% for the Cannon Lake 10nm CPU, well below the 60% Intel had expected at this point. "
"We since have confirmed that the graphics is indeed fused off. Intel’s official line is that this processor was released with a specific target market in mind, and it fulfils the role required. What exactly this market is, and at what price point, is still a mystery, even in 2019. However some analysts believe the graphics was a dodo out of the door due to uneconomically viable yields, as well as this chip not making any sense commercially for the product segment it ended up in – it was put into the market just to fulfil a promise to investors.
Moving from May to August, and no new processors or devices relating to Intel’s 10nm were announced. However Charlie over at SemiAccurate published knowledge that there were issues with Intel’s 10nm process as currently presented. He reports that Intel’s yields were sub-10% for the Cannon Lake 10nm CPU, well below the 60% Intel had expected at this point. "
And so? What you want to aim at? Intel had a big problem with their 10nm mostly because they tried to go beyond what was the actual limits of the technology they were using. But they have the money for that and also to develop the next PP with EUV in parallel.
GF failed the 20nm shrink, could not implement the FinFet tech by their own which then they bought from Samsung and failed to develop the 7nm node. Since they became an independent foundry from AMD they manage to create the 12nm PP as a refinement of the 14nm. That is, they have proved not being able to develop new PP at all.
Of course this has got NOTHING to do with 7nm development. 7nm has always taken place on 300mm wafers. This fab in the process of being sold, is a 200mm fab!
It's probably the other way around: GlobalFoundries has been hard up for cash, and cancelling 7nm (which would have cost billions) and selling a part of its business are all reflections of that.
GlobalFoundries has been hurting for a LONG time - AMD spun it off partly because they couldn't afford for it to keep on losing money, IBM sold its business to GF for similar reasons, and GF was purchased at a fairly low price partly because the business model hasn't looked sustainable. Chances are that GF lost money on 14/12nm development, and couldn't stand to lose that amount on 7nm (and even if they could.... couldn't do ti after that).
To add a little more context, VIS was founded by ex TSMC CEO Morris Chang. So the two are closely related.
And ~$240M for Fab and IP as well as all the business is pretty much a bargain. And MEMS is still a huge opportunity with no clear winner yet. ( Well technically STMicro but I don't have much faith in them competing )
I'm pretty impressed, it's only taken the Saudis what... 10 years? to run what was AMD's fab division into the ground. Give it another decade for GloFo to cease existence.
Don't count your chickens before they are hatched. The fab business is changing, costs have exploded and there's going to be a massive slowdown and consolidation. There's only going to be at best 2 companies that can get the money to build a $20billion dollar fab. That's what 5nm is expected to cost and costs are exponential growth with EUV and above.
It's very possible that GloFo made the smartest move they could have made.
Agreed - this is the only route forward that makes sense. Their closest competitor (TSMC) brings in 6x their revenue, Samsung Semi apparently nets 10x that again. There is simply no way they can match the investment of competitors of that size.
This move is the precise opposite of running something into the ground. Adaptation is key to survival; if you can't outgrow your competitors then you have to specialize.
It is the UAE (United Arab Emirates) or more specifically the Abu Dhabi Emirate with its SWF Mubadala. Not the Saudis. With that being said, it has brought down Chartered Semi, IBM Micro also with it
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.
34 Comments
Back to Article
baka_toroi - Friday, February 1, 2019 - link
"GlobalFoundries will get $236 million and will invest them in other 200-mm fabs..."Is that 200 nm or what?
peevee - Friday, February 1, 2019 - link
It is the diameter of a wafer.Mikewind Dale - Friday, February 1, 2019 - link
His question is whether it's nanometers or millimeters.shabby - Friday, February 1, 2019 - link
Diameter of a water is in mm.shabby - Friday, February 1, 2019 - link
Wafer.Cooe - Friday, February 1, 2019 - link
And it should be obvious from peevee's answer it's the latter. 200nm wafers? Nobody's that dumb, if peevee needed to spell that last bit out for you, you shouldn't be on AnandTech.baka_toroi - Friday, February 1, 2019 - link
Protozoans also need wafers!wr3zzz - Friday, February 1, 2019 - link
200 mm is 8" wafer.HardwareDufus - Friday, February 1, 2019 - link
maybe you should be a little nicer. there was no need to insinuate the poster was dumb.shabby - Friday, February 1, 2019 - link
No one insinuated the op was dumb, just Mike wind.HStewart - Friday, February 1, 2019 - link
I wonder how much this has to do with stopping it 7nm development. That development may have hurt GlobalFoundries a lot.CiccioB - Friday, February 1, 2019 - link
What many do not understood is that 7nm is a very expensive node both for R&D and for production point of view.GF simply can't sustain that expenses and for such is now relegated as foundry playing in B league.
For me in 5 years GF will be dead or sold to a bigger foundry that may be interested in their buildings, not their technology or it will play in very niche market (RF, low power cheap IC, anything that is CHEAP) where it won't be able to come out anymore to fab for high performance industry unless a complete change in materials and cheaper machinery will come.
In few years will be left with only 3 fabs capable of using latest PPs: TSMC, Intel and Samsung, and I'm not sure there will be a place in the market for all three of them.
FreckledTrout - Friday, February 1, 2019 - link
Agree and I do see a place for all three fabs going forward. I wonder if they will start sharing IP once we see 4-5nm as its going to even more expensive than 7nm.sing_electric - Monday, February 4, 2019 - link
That was my thought, too, that we're likely to see more extensive sharing of IP between companies as we shrink processes further. That's not the worst thing in the world, especially if it helps keep all 3 in the market. Every process shrink from 180nm down to 7nm has resulted in at least one foundry dropping out of the "bleeding edge" market (with the exception of the 22-14nm shrink), and it's bad enough there's only 2 choices for a fabless company (like AMD, Nvidia, etc.) to place orders for the latest process.0ldman79 - Friday, February 1, 2019 - link
You've got to wonder.I do think that lithography is coming to a point of diminished returns, so 12nm and above might work well for SSD, I/O, wireless, NIC, memory, etc...
Eventually they'll have to go to the next step otherwise they're done for. Everyone will settle in at 7nm, 5nm or 3nm, but we're still at the end of process improvements moving the industry forward.
flgt - Friday, February 1, 2019 - link
There is a ton of money in RF, ADC/DAC’s,power management. There is probably 10X more revenue in those areas for the products I work on than the processor, so no reason they won’t be around awhile. The key is just to know what you want to be and focus on it. Seems like that’s what they’re doing.CiccioB - Monday, February 4, 2019 - link
10x revenue but 1/10 net profits.It is not by chance that TSMC started ramping up their PP investments only after ARM SoC using latest available PP boom happened.
If you want to make real money you have to produce with the latest PP. Otherwise you are a general foundry like any other (and you may know that all foundris, TSMC, Intel and Samsung operate in the same RF and network, which is another big market, as well).
Declaring their debacle on the new PP, GF simply puts itself in the second lines of foundries on the global market and with diminishing profits (while being a quite large company) it probably will further go behind to 3rd ot 4th line.
In 5 years they be none you will know bout as a consumer and will probably be swallowed (or will merge with) other small foundries to try to create a better marketing scale value, that is essentially what makes foundries make money.
sing_electric - Monday, February 4, 2019 - link
The death of companies making latest process silicon really is striking. (Here's a chart going from 130nm to 22nm): http://www.fabtech.org/news/_a/bleading-edge_found...With the exception of 14nm, EVERY process shrink has kicked at least one competitor out of the game, and this is NOT good for consumers.
We're in a similar situation with say, DRAM production, where 3 suppliers now contorl the market.
frenchy_2001 - Friday, February 1, 2019 - link
They stopped development of 7nm and beyond because costs are prohibitive and they could not compete.Their execution had been lacking for the past few years, especially in process.
They had to *license* Samsung's 14nm process, instead of using their own, as planned.
Process is only getting more expensive and competition is moving faster. They retreated on their own, differentiated processes. What nobody else offers.
This is rather smart. They have a market to serve and the competence to do so, with less competition. Their future, 5~10 years, is assured. Beyond that, they may finally push for smaller sizes at a much lower cost (and similarly add their own twist to it, like on their 12nm processes).
There is a place for such niche manufacturing, even if it is less glamorous.
qcmadness - Friday, February 1, 2019 - link
Intel has poured billions into 10nm and what do they get? <10% yield for a dual-core product.peevee - Saturday, February 2, 2019 - link
How do you know what their yield is?qcmadness - Sunday, February 3, 2019 - link
https://www.anandtech.com/show/13405/intel-10nm-ca..."We since have confirmed that the graphics is indeed fused off. Intel’s official line is that this processor was released with a specific target market in mind, and it fulfils the role required. What exactly this market is, and at what price point, is still a mystery, even in 2019. However some analysts believe the graphics was a dodo out of the door due to uneconomically viable yields, as well as this chip not making any sense commercially for the product segment it ended up in – it was put into the market just to fulfil a promise to investors.
Moving from May to August, and no new processors or devices relating to Intel’s 10nm were announced. However Charlie over at SemiAccurate published knowledge that there were issues with Intel’s 10nm process as currently presented. He reports that Intel’s yields were sub-10% for the Cannon Lake 10nm CPU, well below the 60% Intel had expected at this point. "
qcmadness - Sunday, February 3, 2019 - link
https://www.anandtech.com/show/13405/intel-10nm-ca..."We since have confirmed that the graphics is indeed fused off. Intel’s official line is that this processor was released with a specific target market in mind, and it fulfils the role required. What exactly this market is, and at what price point, is still a mystery, even in 2019. However some analysts believe the graphics was a dodo out of the door due to uneconomically viable yields, as well as this chip not making any sense commercially for the product segment it ended up in – it was put into the market just to fulfil a promise to investors.
Moving from May to August, and no new processors or devices relating to Intel’s 10nm were announced. However Charlie over at SemiAccurate published knowledge that there were issues with Intel’s 10nm process as currently presented. He reports that Intel’s yields were sub-10% for the Cannon Lake 10nm CPU, well below the 60% Intel had expected at this point. "
CiccioB - Monday, February 4, 2019 - link
And so? What you want to aim at?Intel had a big problem with their 10nm mostly because they tried to go beyond what was the actual limits of the technology they were using.
But they have the money for that and also to develop the next PP with EUV in parallel.
GF failed the 20nm shrink, could not implement the FinFet tech by their own which then they bought from Samsung and failed to develop the 7nm node.
Since they became an independent foundry from AMD they manage to create the 12nm PP as a refinement of the 14nm. That is, they have proved not being able to develop new PP at all.
happily1986 - Saturday, February 2, 2019 - link
Of course this has got NOTHING to do with 7nm development. 7nm has always taken place on 300mm wafers. This fab in the process of being sold, is a 200mm fab!sing_electric - Monday, February 4, 2019 - link
It's probably the other way around: GlobalFoundries has been hard up for cash, and cancelling 7nm (which would have cost billions) and selling a part of its business are all reflections of that.GlobalFoundries has been hurting for a LONG time - AMD spun it off partly because they couldn't afford for it to keep on losing money, IBM sold its business to GF for similar reasons, and GF was purchased at a fairly low price partly because the business model hasn't looked sustainable. Chances are that GF lost money on 14/12nm development, and couldn't stand to lose that amount on 7nm (and even if they could.... couldn't do ti after that).
ksec - Friday, February 1, 2019 - link
To add a little more context, VIS was founded by ex TSMC CEO Morris Chang. So the two are closely related.And ~$240M for Fab and IP as well as all the business is pretty much a bargain. And MEMS is still a huge opportunity with no clear winner yet. ( Well technically STMicro but I don't have much faith in them competing )
londedoganet - Sunday, February 3, 2019 - link
GloFo and TSMC were never the same company. How are they “closely related”?KateH - Sunday, February 3, 2019 - link
i think they meant VIS and TSMC are closely related due to the founder of VIS being a former TSMC execThe_Assimilator - Saturday, February 2, 2019 - link
I'm pretty impressed, it's only taken the Saudis what... 10 years? to run what was AMD's fab division into the ground. Give it another decade for GloFo to cease existence.rahvin - Saturday, February 2, 2019 - link
Don't count your chickens before they are hatched. The fab business is changing, costs have exploded and there's going to be a massive slowdown and consolidation. There's only going to be at best 2 companies that can get the money to build a $20billion dollar fab. That's what 5nm is expected to cost and costs are exponential growth with EUV and above.It's very possible that GloFo made the smartest move they could have made.
Spunjji - Monday, February 4, 2019 - link
Agreed - this is the only route forward that makes sense. Their closest competitor (TSMC) brings in 6x their revenue, Samsung Semi apparently nets 10x that again. There is simply no way they can match the investment of competitors of that size.This move is the precise opposite of running something into the ground. Adaptation is key to survival; if you can't outgrow your competitors then you have to specialize.
Lonyo - Sunday, February 3, 2019 - link
Because AMD were doing such a great job with it, that's why they were trying so hard to hang on to it...rocketbuddha - Sunday, February 3, 2019 - link
It is the UAE (United Arab Emirates) or more specifically the Abu Dhabi Emirate with its SWF Mubadala. Not the Saudis.With that being said, it has brought down Chartered Semi, IBM Micro also with it