Speculation: Intel will become fabless

Page 11 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

With the loss of its manufacturing lead, will Intel become fabless?

  • Yes, Intel is a product designer at heart, and they will seek a more flexible fabless model.

    Votes: 19 13.9%
  • No, manufacturing is integral to Intel, and they will continue to invest to stay competitive.

    Votes: 118 86.1%

  • Total voters
    137

Doug S

Senior member
Feb 8, 2020
250
301
96
What's stopping them from moving to TSMC 5nm? The product has insignificant volume so competition for wafers shouldn't be a concern
Yes I think if there was a "Intel 7nm vs TSMC" decision to be made, especially for a product slated for late 2021, that TSMC's 5nm would be the only choice because the density of the two and performance should be roughly comparable. TSMC's 7nm would be unworkable with a ~2x density difference.

The competition for 5nm wafers wouldn't be so fierce by that time, as it would be fully ramped up. The bulk of Apple's 5nm run would be behind them at that point, and customers like them who following TSMC's leading edge would be taping out N3 designs and maybe getting back A0 silicon around that time.
 

Vattila

Senior member
Oct 22, 2004
505
468
136
While the PC market has been shrinking, it is only shrinking on the client side. Servers - which are much larger chips - are still growing, and they've developed a presence in HPC and own the majority of cloud.
Thanks, that is enlightening. I have incorrectly been thinking about it mainly in terms of shipped units. But as you point out, server chips are much larger, and hence also yield much worse, than small client chips going into notebooks, hence require more wafer starts relative to finished working chips. Still, despite a declining PC market and growing server market, there is a vast difference in units going into x86 client and server. Am I right in estimating it to around 260 million vs ~20 million per year? That is an order of magnitude difference. So, I would presume more fab capacity is needed to serve the client market than the server market, unless a server chip consumes 10 times more wafer capacity than a client chip on average.

Correct me if I am wrong.
 

Doug S

Senior member
Feb 8, 2020
250
301
96
Thanks, that is enlightening. I have incorrectly been thinking about it mainly in terms of shipped units. But as you point out, server chips are much larger, and hence also yield much worse, than small client chips going into notebooks, hence require more wafer starts relative to finished working chips. Still, despite a declining PC market and growing server market, there is a vast difference in units going into x86 client and server. Am I right in estimating it to around 260 million vs ~20 million per year? That is an order of magnitude difference. So, I would presume more fab capacity is needed to serve the client market than the server market, unless a server chip consumes 10 times more wafer capacity than a client chip on average.

Correct me if I am wrong.
Well don't forget that many servers have more than one socket, and the biggest cloud providers like MIcrosoft, Amazon, Alibaba etc. are the market equivalent of "dark matter" where we don't know what they're doing since they buy components and build their own servers. Figures from companies like IDC don't include them.

Maybe Intel has released info in the recent past that lets us determine how many server CPUs they've sold in a given quarter/year, i.e. have they ever provided revenue figures and an ASP that lets us do the math?

Something like Intel's 56 core Xeon actually consumes MORE than 10x the fab capacity, even if yields were the same as client CPUs. It is two 694 mm^2 dies or around 1400 mm^2, while the average client die size is probably around 100 mm^2 (before you point out all the examples of client CPUs that are bigger, remember how much of the market (even moreso outside the US/EU) is low end stuff with "Pentium" in the name)

Though I would imagine when you average out all the biggest ones like that 56 core Xeon and all the lower end server CPUs with a lot fewer cores the average server CPU is probably more like 500 mm^2 or 5x the wafer capacity, maybe 6x if you account for the reduced yields. So not 10x but if you add one server CPU sales for every 5 or 6 client CPU sales you lose you break even. Compare the drop in the client PC market with the rise in the server market over the past decade, what's that tell you?
 
  • Like
Reactions: Tlh97 and Vattila

Hitman928

Platinum Member
Apr 15, 2012
2,588
1,813
136
Well don't forget that many servers have more than one socket, and the biggest cloud providers like MIcrosoft, Amazon, Alibaba etc. are the market equivalent of "dark matter" where we don't know what they're doing since they buy components and build their own servers. Figures from companies like IDC don't include them.

Maybe Intel has released info in the recent past that lets us determine how many server CPUs they've sold in a given quarter/year, i.e. have they ever provided revenue figures and an ASP that lets us do the math?

Something like Intel's 56 core Xeon actually consumes MORE than 10x the fab capacity, even if yields were the same as client CPUs. It is two 694 mm^2 dies or around 1400 mm^2, while the average client die size is probably around 100 mm^2 (before you point out all the examples of client CPUs that are bigger, remember how much of the market (even moreso outside the US/EU) is low end stuff with "Pentium" in the name)

Though I would imagine when you average out all the biggest ones like that 56 core Xeon and all the lower end server CPUs with a lot fewer cores the average server CPU is probably more like 500 mm^2 or 5x the wafer capacity, maybe 6x if you account for the reduced yields. So not 10x but if you add one server CPU sales for every 5 or 6 client CPU sales you lose you break even. Compare the drop in the client PC market with the rise in the server market over the past decade, what's that tell you?
They don't provide ASP except as a % change but PC revenue last quarter was $9.5B and Data Center was $7.1B. It's really hard to say how many wafers each takes because we don't know what kind of discounts Intel is giving for their server chips now and it also seems like the 10 nm yields are still really bad for the PC group and we also don't know what kind of discounts Intel is giving OEMs in the space either. With that said, I would definitely lean towards the PC groups still taking up the majority of wafers given typically how much higher margins are for server chips.
 
  • Like
Reactions: Tlh97 and Vattila

DrMrLordX

Lifer
Apr 27, 2000
16,022
4,970
136
On the likelihood that Intel could catch up to AMD's Genoa chip, scheduled to arrive before the end of next year, by porting their own server chip design to TSMC's N5, I am doubtful.
Intel has more bargaining chips available than AMD. They might not be able to outperform AMD, but they can out-volume them.
 
  • Like
Reactions: Tlh97 and Vattila

scannall

Golden Member
Jan 1, 2012
1,674
1,037
136
Intel has more bargaining chips available than AMD. They might not be able to outperform AMD, but they can out-volume them.
From the (sketchy) details I can eek out, TSMC has about 3 times the fab capacity that Intel does. 800,000 WPM vs. 2,400,000.
 

DrMrLordX

Lifer
Apr 27, 2000
16,022
4,970
136
From the (sketchy) details I can eek out, TSMC has about 3 times the fab capacity that Intel does. 800,000 WPM vs. 2,400,000.
AMD doesn't have access to all that capacity. I doubt AMD can convince TSMC to provide them with 800k WPM.
 

scannall

Golden Member
Jan 1, 2012
1,674
1,037
136
AMD doesn't have access to all that capacity. I doubt AMD can convince TSMC to provide them with 800k WPM.
No, probably not. Just some posters seem to think that TSMC is some backyard operation that doesn't produce much. But with chiplets they don't need 800k either. Is Intel a 'fab monster' for volume. Yes, they are. But they aren't the biggest, or even close.
 
  • Like
Reactions: Tlh97

arandomguy

Senior member
Sep 3, 2013
524
138
116
Scale up and scale down in this industry is much slower though. The DDR4 price ramp that more than doubled costs was only the result of a <10% in-balance with demand over supply. The problem with an Intel move over significantly to TSMC is a possibly MAD scenario (for everyone, not just AMD/Intel, all related consumers and producers, well except TSMC) in terms of the demand to supply shift.

I know in the enthusiast PC world people want to frame everything as Intel vs AMD but to me the real problem for Intel over the last few years is they've been losing the battle against TSMC. It's really an Intel vs TSMC situation that TSMC has been winning.
 

moinmoin

Golden Member
Jun 1, 2017
1,657
1,594
106
Scale up and scale down in this industry is much slower though. The DDR4 price ramp that more than doubled costs was only the result of a <10% in-balance with demand over supply. The problem with an Intel move over significantly to TSMC is a possibly MAD scenario (for everyone, not just AMD/Intel, all related consumers and producers, well except TSMC) in terms of the demand to supply shift.

I know in the enthusiast PC world people want to frame everything as Intel vs AMD but to me the real problem for Intel over the last few years is they've been losing the battle against TSMC. It's really an Intel vs TSMC situation that TSMC has been winning.
Indeed. Due to not competing directly or even in the same wider markets until recently Intel was able to essentially ignore TSMC as a peripheral figure. This completely changed when AMD switched to TSMC with Zen 2. Now TSMC's progress suddenly directly affects Intel's core markets. I often criticize @DrMrLordX for having too high expectations in this regard, but Intel should consider itself lucky AMD is indeed not a bleeding edge node customer of TSMC.
 
  • Like
Reactions: Tlh97

jpiniero

Diamond Member
Oct 1, 2010
7,924
1,206
126
They don't provide ASP except as a % change but PC revenue last quarter was $9.5B and Data Center was $7.1B. It's really hard to say how many wafers each takes because we don't know what kind of discounts Intel is giving for their server chips now
The Intel 10-Q has the revenue impact of the ASP and volume change. Volume increasing by 29% added 1.3B in revenue; ASP increasing by 5% added 302M in revenue.
 

senttoschool

Golden Member
Jan 30, 2010
1,412
153
106
Indeed. Due to not competing directly or even in the same wider markets until recently Intel was able to essentially ignore TSMC as a peripheral figure. This completely changed when AMD switched to TSMC with Zen 2. Now TSMC's progress suddenly directly affects Intel's core markets. I often criticize @DrMrLordX for having too high expectations in this regard, but Intel should consider itself lucky AMD is indeed not a bleeding edge node customer of TSMC.
This is true, for now. If AMD continues to grow in both CPU and GPU volume, it can become just as important as Apple is to TSMC. So in the near future, it's quite reasonable to expect AMD and Apple to both share the bleeding edge node.

Thus, if Intel does ever threaten AMD by having an equivalent node, AMD has an ace card in its sleeves: use the absolute newest TSMC node to buy itself time and stay ahead.

I bet this is playing into Intel's new openness to 3rd party fabs - that AMD will likely always be 1-2 nodes ahead no matter what.

This is why I believe Intel switching to TSMC for their server, laptop, and desktop CPUs is a disaster for AMD. It completely removes AMD's node advantage and significantly increases wafer costs for AMD.

I think this is what will happen. Intel will sign an agreement with TSMC to manufacture all their CPUs there. As part of the agreement, TSMC will force Intel to exit the bleeding node competition. Intel will continue to manufacture their modems and SSDs on their own fabs but those don't require the newest nodes. And slowly, Intel will shed its fabs and become fabless.

As for TSMC, they will essentially become the Intel of the last three decades: a monopoly. And they gain a huge customer at the same time.

I would load up on TSMC stocks. I also expect AMD stocks to tank if Intel comes to an agreement with TSMC.

This is why I laugh at AMD's stock gaining 16% on Friday and Intel dropping 16%. Yes, it's really bad for Intel but it's bad for AMD too. The big winner is TSMC.
 
Last edited:

maddie

Diamond Member
Jul 18, 2010
3,159
1,829
136
This is true, for now. If AMD continues to grow in both CPU and GPU volume, it can become just as important as Apple is to TSMC. So in the near future, it's quite reasonable to expect AMD and Apple to both share the bleeding edge node.

Thus, if Intel does ever threaten AMD by having an equivalent node, AMD has an ace card in its sleeves: use the absolute newest TSMC node to buy itself time and stay ahead.

I bet this is playing into Intel's new openness to 3rd party fabs - that AMD will likely always be 1-2 nodes ahead no matter what.

This is why I believe Intel switching to TSMC for their server, laptop, and desktop CPUs is a disaster for AMD. It completely removes AMD's node advantage and significantly increases wafer costs for AMD.

I think this is what will happen. Intel will sign an agreement with TSMC to manufacture all their CPUs there. As part of the agreement, TSMC will force Intel to exit the bleeding node competition. Intel will continue to manufacture their modems and SSDs on their own fabs but those don't require the newest nodes. And slowly, Intel will shed its fabs and become fabless.

As for TSMC, they will essentially become the Intel of the last three decades: a monopoly. And they gain a huge customer at the same time.

I would load up on TSMC stocks. I also expect AMD stocks to tank if Intel comes to an agreement with TSMC.

This is why I laugh at AMD's stock gaining 16% on Friday and Intel dropping 16%. Yes, it's really bad for Intel but it's bad for AMD too. The big winner is TSMC.
I think AMD presently is 'at least' just as important to TSMC as Apple.

A scenario can be presented. In any case, TSMC wants to dominate semi production, especially in the high end.

Apple stays a numerically smaller player % wise in # world shipped units. They keep margins up and go for high prices/premium products. In this case, with their recent announcement of an ARM transition to most products, they will continue to be the 1st on bleeding edge, limited volume nodes. No room for AMD who will follow closely after Apple moves on.

So we have Intel being squeezed by AMD and Apple + AMD being squeezed by Apple to a lesser degree.

Stock prices are the shortest of time scales. For the next few years, there is little Intel can do. Time can't be jumped. AMD is a big winner here, also TSMC.
 
  • Like
Reactions: Tlh97

arandomguy

Senior member
Sep 3, 2013
524
138
116
As for TSMC, they will essentially become the Intel of the last three decades: a monopoly. And they gain a huge customer at the same time.

I would load up on TSMC stocks. I also expect AMD stocks to tank if Intel comes to an agreement with TSMC.

This is why I laugh at AMD's stock gaining 16% on Friday and Intel dropping 16%. Yes, it's really bad for Intel but it's bad for AMD too. The big winner is TSMC.
While Intel did have the technology lead as a fab I don't think you could ever consider them a monopoly as a fab in the business sense as they were effectively a non competitor in that space. You could argue in hindsight this was likely their biggest mistake in that they never leveraged that advantage into an actual dominant business position, instead only being focused on the x86 CPU business.

In this hypothetical scenario it's going to create a single source for all customers in the semi industry assuming Samsung also falls as well compounded by a snowball in investment to TSMC.
 

senttoschool

Golden Member
Jan 30, 2010
1,412
153
106
While Intel did have the technology lead as a fab I don't think you could ever consider them a monopoly as a fab in the business sense as they were effectively a non competitor in that space. You could argue in hindsight this was likely their biggest mistake in that they never leveraged that advantage into an actual dominant business position, instead only being focused on the x86 CPU business.
Precisely. Intel was a monopoly in terms of bleeding edge.

Stock prices are the shortest of time scales. For the next few years, there is little Intel can do. Time can't be jumped. AMD is a big winner here, also TSMC.
I think the stock market has reacted incorrectly to the news that Intel might switch to 3rd party fabs. Intel's stock should drop, rightfully so. But AMD's stock should also drop. I'm betting that AMD stocks will not perform well in the next 3 years once people realize that Intel moving to TSMC is simply not a good thing for AMD. Meanwhile, Intel will lower prices to minimize market share loss.

AMD's stock was already pricing in perfection years ahead as its P/E was 150. Intel's P/E at the moment is only 10.

Basically, I'm betting that Intel will move to TSMC and stop the bleeding faster than AMD can gain meaningful market share away from Intel. And once it does, Intel should be able to outcompete AMD for TSMC wafers. Both Intel and AMD will see a decline in gross margins if they have to depend on TSMC. Thus, the only winner here is TSMC.
 
Last edited:

ondma

Golden Member
Mar 18, 2018
1,279
324
106
Precisely. Intel was a monopoly in terms of bleeding edge.


I think the stock market has reacted incorrectly to the news that Intel might switch to 3rd party fabs. Intel's stock should drop, rightfully so. But AMD's stock should also drop. I'm betting that AMD stocks will not perform well in the next 3 years once people realize that Intel moving to TSMC is simply not a good thing for AMD. Meanwhile, Intel will lower prices to minimize market share loss.

AMD's stock was already pricing in perfection years ahead as its P/E was 150. Intel's P/E at the moment is only 10.

Basically, I'm betting that Intel will move to TSMC and stop the bleeding faster than AMD can gain meaningful market share away from Intel.
Interesting take. I still am not convinced that Intel will (or even would be able to in a reasonable time) shift the majority of their production to TSMC.
For sure I agree the price of AMD is through the roof, and probably over valued. Of course I have felt this way for a while now, and it just keeps going up.
 

senttoschool

Golden Member
Jan 30, 2010
1,412
153
106
Interesting take. I still am not convinced that Intel will (or even would be able to in a reasonable time) shift the majority of their production to TSMC.
For sure I agree the price of AMD is through the roof, and probably over valued. Of course I have felt this way for a while now, and it just keeps going up.
I owned a lot of AMD stocks before Zen2 and exited before the recent all-time highs.

I think AMD's stock is being powered by a lot of fanboys and a lot of institutional investors that aren't as well-tuned to tech trends as enthusiasts.

AMD has some fundamental problems:
  • For CPUs, Intel moving to TSMC will erode its node advantage and cause a decrease in gross margins since they will have to pay more for wafers.
  • The x86 market is not growing. It will shrink.
  • In the hyperscalers space, Amazon has already demonstrated that its ARM Graviton CPU can compete against x86. You can bet that Amazon will eventually want to switch to primarily using its in-house CPUs. Microsoft and Google will likely follow this trend.
  • For AI, Nvidia has built incredible moat in the industry with CUDA that AMD is just not likely to ever compete with its own compute GPUs
  • For gaming, Nvidia has time and time again shown that not only do they offer higher performance, they're more innovative with things like DLSS and ray tracing.
  • Consoles are a low margin business
  • Mobile gaming is now a bigger business than PC and console gaming. AMD has no access to this market. The market is controlled by Qualcomm and Apple.
In the semiconductor business, it seems like you either become the monopoly or you're at a distant second place. I don't see AMD becoming a monopoly in any of the segments right now. No matter where you look, AMD is competing against players far bigger than they are.

The one clear area AMD can monopolize is cloud game streaming. Google Stadia is built on AMD GPUs. Xbox and Playstation use AMD GPUs so it makes sense that Microsoft and Sony will continue to use AMD GPUs for game streaming tech. But how big is this market really?

So to summarize, AMD's best product (x86 CPUs) will lose its node advantage, is in a declining market, and is competing against big ARM companies in Qualcomm and Amazon. Its GPUs are a distant second in PC gaming and AI. They might have a monopoly in console gaming and cloud gaming, but it's currently low margin and the market size might be small.
 
Last edited:

maddie

Diamond Member
Jul 18, 2010
3,159
1,829
136
Precisely. Intel was a monopoly in terms of bleeding edge.


I think the stock market has reacted incorrectly to the news that Intel might switch to 3rd party fabs. Intel's stock should drop, rightfully so. But AMD's stock should also drop. I'm betting that AMD stocks will not perform well in the next 3 years once people realize that Intel moving to TSMC is simply not a good thing for AMD. Meanwhile, Intel will lower prices to minimize market share loss.

AMD's stock was already pricing in perfection years ahead as its P/E was 150. Intel's P/E at the moment is only 10.

Basically, I'm betting that Intel will move to TSMC and stop the bleeding faster than AMD can gain meaningful market share away from Intel. And once it does, Intel should be able to outcompete AMD for TSMC wafers. Both Intel and AMD will see a decline in gross margins if they have to depend on TSMC. Thus, the only winner here is TSMC.
If Intel shifts to TSMC or any other fab, it won't be the Intel of today. You're trying to predict a change by assuming the effects will only be superficial, eg using another fab for your product, but all else is as before.
 

A///

Senior member
Feb 24, 2017
201
75
71
No, probably not. Just some posters seem to think that TSMC is some backyard operation that doesn't produce much. But with chiplets they don't need 800k either. Is Intel a 'fab monster' for volume. Yes, they are. But they aren't the biggest, or even close.
Blind support for "your team" is never a good look for anyone, unless dijon has joined the forum and posts here.
 

Hitman928

Platinum Member
Apr 15, 2012
2,588
1,813
136
If Intel shifts to TSMC or any other fab, it won't be the Intel of today. You're trying to predict a change by assuming the effects will only be superficial, eg using another fab for your product, but all else is as before.
Not only that but I highly doubt TSMC gives Intel any kind of priority status as long as Intel is continuing their fab efforts. It seems like people thinking buying wafers is like a bidding process where more money buys you exclusive access. There's some truth to that but its not nearly so simple. When you're talking about fabs and these very large customers, it becomes more of a partnership and the fab is going to analyze not only how many wafers you are requesting today but also what you are likely to be buying in the next 2, 5, 10 years.

TSMC isn't going to want to be Intel's crutch until they figure out their own process and then go back to their own fabs. TSMC also isn't likely to prioritize Intel over existing long term customers which means that they would need to build capacity out to accommodate Intel, but again, they're not going to do that while there's a chance that Intel goes back to their own fabs, not without a huge long term financial commitment from Intel which I don't see Intel committing to unless they really are giving up on their own fabs. If Intel does give up on their own fabs, you're looking at this happening at least 3 years down the road and Intel will go through a very large restructure both in terms of personnel as well as business model. As you said, it won't be the Intel we know today and the competitive landscape will most likely be very different as well and most likely not in Intel's favor.
 

eek2121

Senior member
Aug 2, 2005
445
333
136
I owned a lot of AMD stocks before Zen2 and exited before the recent all-time highs.

I think AMD's stock is being powered by a lot of fanboys and a lot of institutional investors that aren't as well-tuned to tech trends as enthusiasts.

AMD has some fundamental problems:
  • For CPUs, Intel moving to TSMC will erode its node advantage and cause a decrease in gross margins since they will have to pay more for wafers.
  • The x86 market is not growing. It will shrink.
  • In the hyperscalers space, Amazon has already demonstrated that its ARM Graviton CPU can compete against x86. You can bet that Amazon will eventually want to switch to primarily using its in-house CPUs. Microsoft and Google will likely follow this trend.
  • For AI, Nvidia has built incredible moat in the industry with CUDA that AMD is just not likely to ever compete with its own compute GPUs
  • For gaming, Nvidia has time and time again shown that not only do they offer higher performance, they're more innovative with things like DLSS and ray tracing.
  • Consoles are a low margin business
  • Mobile gaming is now a bigger business than PC and console gaming. AMD has no access to this market. The market is controlled by Qualcomm and Apple.
In the semiconductor business, it seems like you either become the monopoly or you're at a distant second place. I don't see AMD becoming a monopoly in any of the segments right now. No matter where you look, AMD is competing against players far bigger than they are.

The one clear area AMD can monopolize is cloud game streaming. Google Stadia is built on AMD GPUs. Xbox and Playstation use AMD GPUs so it makes sense that Microsoft and Sony will continue to use AMD GPUs for game streaming tech. But how big is this market really?

So to summarize, AMD's best product (x86 CPUs) will lose its node advantage, is in a declining market, and is competing against big ARM companies in Qualcomm and Amazon. Its GPUs are a distant second in PC gaming and AI. They might have a monopoly in console gaming and cloud gaming, but it's currently low margin and the market size might be small.
...and the “doom & gloom”, “x86 is dead”, and “ARM will win” continues on this forum.

Meanwhile Intel continues to sell more chips than they can make, NVIDIA may buy out ARM, and Apple has been avoiding using the term “ARM” to describe their chips.

What I think will happen is:

Intel will use TSMC or Samsung for their GPUs and maybe some other trivial stuff. They will keep the rest of it in house. They will only be outsourcing to free up capacity. Apple leaving will certainly cause them a temporary financial hit, but unless other core partners start bailing they will be fine. Once 10nm reaches maturity, Intel won’t have the power efficiency issues they have today. They don’t need 7nm for quite a while, provided 10nm performs to expectations.

NVIDIA will buy out ARM for the patents and talent, then they will try to utilize that expertise to continue to make inroads into high performance computing. They will likely utilize that advantage to also release a mobile SoC of some kind. I think that NVIDIA recognizes they will need their own platform in the future since Intel is jumping into the GPU fray, so I would not put it past them to release a high performance CPU at some point.

AMD will continue being competitive with Intel.

Apple will do their own thing with their own closed platform.

Node shrinks are likely to slow down drastically in the near future.

The real wildcard here is GlobalFoundries. They have to have something in the works. A contingency plan for the day that their current process is no longer profitable.
 

senttoschool

Golden Member
Jan 30, 2010
1,412
153
106
If Intel shifts to TSMC or any other fab, it won't be the Intel of today. You're trying to predict a change by assuming the effects will only be superficial, eg using another fab for your product, but all else is as before.
All else is as before? I don't quite understand your argument if there is one.
 

senttoschool

Golden Member
Jan 30, 2010
1,412
153
106
Not only that but I highly doubt TSMC gives Intel any kind of priority status as long as Intel is continuing their fab efforts. It seems like people thinking buying wafers is like a bidding process where more money buys you exclusive access. There's some truth to that but its not nearly so simple. When you're talking about fabs and these very large customers, it becomes more of a partnership and the fab is going to analyze not only how many wafers you are requesting today but also what you are likely to be buying in the next 2, 5, 10 years.
I think this forum agrees that any Intel shift to TSMC will mean Intel has to exit the bleeding edge node competition. No one is arguing against that.

This is my personal prediction on how it's going to play out:

I think this is what will happen. Intel will sign an agreement with TSMC to manufacture all their [future] CPUs there. As part of the agreement, TSMC will force Intel to exit the bleeding node competition. Intel will continue to manufacture their modems and SSDs on their own fabs but those don't require the newest nodes. And slowly, Intel will shed its fabs and become fabless.
 

Hitman928

Platinum Member
Apr 15, 2012
2,588
1,813
136
...and the “doom & gloom”, “x86 is dead”, and “ARM will win” continues on this forum.

Meanwhile Intel continues to sell more chips than they can make, NVIDIA may buy out ARM, and Apple has been avoiding using the term “ARM” to describe their chips.

What I think will happen is:

Intel will use TSMC or Samsung for their GPUs and maybe some other trivial stuff. They will keep the rest of it in house. They will only be outsourcing to free up capacity. Apple leaving will certainly cause them a temporary financial hit, but unless other core partners start bailing they will be fine. Once 10nm reaches maturity, Intel won’t have the power efficiency issues they have today. They don’t need 7nm for quite a while, provided 10nm performs to expectations.

NVIDIA will buy out ARM for the patents and talent, then they will try to utilize that expertise to continue to make inroads into high performance computing. They will likely utilize that advantage to also release a mobile SoC of some kind. I think that NVIDIA recognizes they will need their own platform in the future since Intel is jumping into the GPU fray, so I would not put it past them to release a high performance CPU at some point.

AMD will continue being competitive with Intel.

Apple will do their own thing with their own closed platform.

Node shrinks are likely to slow down drastically in the near future.

The real wildcard here is GlobalFoundries. They have to have something in the works. A contingency plan for the day that their current process is no longer profitable.
GlobalFoundries is trying to be a leader in different markets now. They are focused more on automotive, sensors, and similar applications where their FD-SOI process has benefits and are trying to build an IP partnership program so that they become kind of a one stop shop for people designing in these spaces. They've made some progress here and are getting more and more attention in this space. In my experience, TSMC has a leg up in terms of customer experience but they don't really specifically target the markets GF is really trying to be a leader in.
 

ASK THE COMMUNITY