• Guest, The rules for the P & N subforum have been updated to prohibit "ad hominem" or personal attacks against other posters. See the full details in the post "Politics and News Rules & Guidelines."

Speculation: Intel will become fabless

Page 2 - 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: 20 13.0%
  • No, manufacturing is integral to Intel, and they will continue to invest to stay competitive.

    Votes: 134 87.0%

  • Total voters
    154

Vattila

Senior member
Oct 22, 2004
568
645
136
They might open up their fabs to third parties in a big way, but why would they go fabless?
Thanks for a well argued, on-topic reply. Yours deserve a point-by-point reply.

Basically, Intel will go fabless because the huge burden of maintaining a lead in process tech is becoming a liability. That's the gist of my argument, which you probably will find better articulated by articles, blog posts and forum discussions on this topic floating around on the Internet.

AMD/GloFo was a terrible long-term deal for AMD.
What about the IBM-GlobalFoundries deal? Is that a bad deal?

The specific terms of the WSA between AMD and GlobalFoundries may initially have been somewhat balanced against AMD. However, I think Lisa Su has fixed that in the latest renegotiation. Also, GlobalFoundries' missteps on 14nm (which led to the subsequent licensing of Samsung's 14LPP process) colours many observers' view on the matter, I guess. Hopefully, GlobalFoundries will redeem themselves with a great home-brewed 7LP process. Just think about it. The latter may beat Intel to the market. Intel falling behind has been unheard of throughout silicon manufacture history. The WSA may turn out to be very good for AMD, primarily because GlobalFoundries' planning and roadmap focus on satisfying it. It is what Lisa Su calls "deep partnerships".

That is not a model to follow, so much as Samsung's business model is.
Samsung is itself considering divesting its fabs. The problem is that not being a pure-play foundry works against them in competition with TSMC's pure-play business model. Customers of TSMC never have to worry about partnering with a competitor (ref. Apple vs Samsung).

I guess this is also the reason why Intel's foundry play has so far failed. They primarily plan the manufacturing roadmap around their own products, not the foundry customer's needs. Customers rather go with TSMC.

[Intel's] business is very much based on having massive production capacity
True. In an agreement with GlobalFoundries (or whoever they sell to), they will have to ensure supply through the WSA (ref. the IBM-GlobalFoundries deal). Still, by going fabless they have access to the capacity of the whole foundry market, which adds flexibility and reduces risk. We can see this play out now for AMD, with Lisa Su dual-sourcing 7nm wafers from TSMC and GlobalFoundries. AMD even has a second-source agreement with Samsung on 14LPP, I think (unless it has run out by now).

True, but that's affecting other companies, as well.
And most design houses have gone fabless for this reason, i.e. the mounting cost of leading-edge semiconductor manufacture.

I definitely think fabless is Intel's destiny. If not in the next couple of generations, then in the end.
 

Cerb

Elite Member
Aug 26, 2000
17,485
33
86
Thanks for a well argued, on-topic reply. Yours deserve a point-by-point reply.

Basically, Intel will go fabless because the huge burden of maintaining a lead in process tech is becoming a liability. That's the gist of my argument, which you probably will find better articulated by articles, blog posts and forum discussions on this topic floating around on the Internet.
Their process, and being able to work on it with the chips, is a large part of them being a behemoth, though, and there is no long-term advantage to getting rid of that. It's a liability as far as the costs of maintaining a manufacturing lead are eating into income, which is based in majority on CPU design. In addition, there's a very simple Intelish catch to going fabless: 5-10% gross profit that someone else is making off of them.

What about the IBM-GlobalFoundries deal? Is that a bad deal?
On IBM's side, IBM was winding down production, anyway, and by also reducing their R&D, they don't need to build much. They are a husk of the company they once were, much like HP.

The specific terms of the WSA between AMD and GlobalFoundries may initially have been somewhat balanced against AMD. However, I think Lisa Su has fixed that in the latest renegotiation.
Repaired it, maybe. Like they were before, AMD could have been a large-scale fab-owning partner in GloFo, instead of kinda sorta having a fab or two, to gradually get rid of. Basically, just doing like they did before GloFo, with the compatible process deal they had for IBM & friends. They'd be stuck with debt, as things are still working themselves out for GloFo, but it's not like going for cash flow to keep afloat is a new way for AMD to operate. But, once they got decent chips made, and could operate their fabs at higher utilization, they'd be left with just that much more revenue, while having the possibility of doing 3rd party fabbing to help buffer the costs during downtimes, yet also having the rest of GloFo to call upon when demand really spiked. AMD went fabless due to bad management, more than anything else. They could have kept their fabs, and also benefited from being part of a consortium, with a better CEO and board.

Samsung is itself considering divesting its fabs. The problem is that not being a pure-play foundry works against them in competition with TSMC's pure-play business model. Customers of TSMC never have to worry about partnering with a competitor (ref. Apple vs Samsung).
This much is true. However, Intel has the opportunity to gradually get there, and work out boundaries along the way. They know what lies ahead, but don't have to do any emergency measures, just yet. They could also just as well fab for competitors or competitors' partners, if it comes down to it. If they make the decision, as the cost vs. revenue problem starts making shareholders think twice, to stick primarily to being a fabricator, and a processor design company second, they can do both. They would make a small amount of money on those chips, and a lot more money on their own.

I guess this is also the reason why Intel's foundry play has so far failed. They primarily plan the manufacturing roadmap around their own products, not the foundry customer's needs. Customers rather go with TSMC.
Of this I have no doubt. While the details of designing and manufacturing chips goes way over my head quite often, I have read from people that did work at Intel or companies that have used their fabs, that it's a royal PITA, compared to the likes of TSMC. Having been able to do their own thing for so long, I suspect that opening up in a big way will take Intel quite some time, and present some major internal hurdles, as they will have to change the way they know how to do things into the ways to do them, or using other tools, that make it easier for third parties.

And most design houses have gone fabless for this reason, i.e. the mounting cost of leading-edge semiconductor manufacture.
Yes, but somebody has to do it, and that is Intel's primary business. The processors are items to move through the fabs, but producing them on a leading edge process, then producing other items on extremely mature processes, has more been a way to fill up those fabs, than it is that their primary business is the CPUs. That is, being able to, at a low cost per unit, produce massive amounts of chips, and sell them for a hefty profit, is Intel's business, much more than the chip design is. Outsourcing that would mean higher costs to produce each unit, and less control over its production. Regardless, however, I think Intel is going to have to change, as time goes on, to a business model working on lower margins. Maybe not 5%, like the Chinese companies often do, but maybe ~25%, or something like that.

I definitely think fabless is Intel's destiny. If not in the next couple of generations, then in the end.
I'm going fabbing being their purpose, and destiny, with x86's dominance just being a happy accident for them along the way.
 
  • Like
Reactions: Pilum and Vattila

Spartak

Senior member
Jul 4, 2015
352
263
136
If they can fix 10nm in a timely fashion and get back in the game with 7nm they will indeed further open their fabs to others for the capital investments needed for the next nodes.

However, if they decide to skip 10nm, and 7nm gets delayed as well they will need to order externally to compete. Competing against 5N GF with their 14nm well into 2021 would be madness.

At that point it would be better to spin off the fabs in the same fashion as AMD did.

Selling to GF will never happen for two reasons: GF lacks the capital for such a huge purchase (IBM was probably over 20x smaller) and also for competitive reasons it seems unlikely Intel would want to buy their chips from AMD's supplier.
 

ehume

Golden Member
Nov 6, 2009
1,511
72
91
Even if they do divest, they will be forced by the government to sell to a US corporation. A spinoff is in essence selling to a US corporation, if (and the government will make the if a rqueirement) it keeps its corporate offices in the US. I think people do not think about the Defense Dept. enough.
 

Vattila

Senior member
Oct 22, 2004
568
645
136
Thanks for all the replies and votes!

The consensus from the poll is "no", Intel will not go fabless.

However, I still think they will need to do something.
  • Spin off the manufacturing division into a fully owned subsidiary, which is run as an independent foundry, or
  • Create a joint venture with e.g. GlobalFoundries or Samsung, retaining majority ownership and control, or
  • Create a joint venture with a minority stake and a wafer supply agreement, or
  • Fully divest the fabs, and secure a wafer supply agreement as part of the deal.
The key, in my view, is to mitigate risk:
  1. Ensure that the design division can do diligent contingency planning, and opt to purchase wafers on the market, if the manufacturing division lags behind.
  2. Operate the manufacturing division as a foundry, catering to customer's needs, as to build a broad customer base, thus ensuring the scale needed to cover the cost of staying on the leading edge, while not being wholly dependent on Intel products to fill the fabs.
This would make the divisions more resilient should the other falter.
 
Last edited:

ehume

Golden Member
Nov 6, 2009
1,511
72
91
Thanks for all the replies and votes!

The consensus from the poll is "no", Intel will not go fabless.

However, I still think they will need to do something.
  • Spin off the manufacturing division into a fully owned subsidiary, which is run as an independent foundry, or
  • Create a joint venture with e.g. GlobalFoundries or Samsung, retaining majority ownership and control, or
  • Create a joint venture with a minority stake and a wafer supply agreement, or
  • Fully divest the fabs, and secure a wafer supply agreement as part of the deal.
The key, in my view, is to mitigate risk:
  1. Ensure that the design division can do diligent contingency planning, and opt to purchase wafers on the market, if the manufacturing division lags behind.
  2. Operate the manufacturing division as a foundry, catering to customer's needs, as to build a broad customer base, thus ensuring the scale needed to cover the cost of staying on the leading edge, while not being wholly dependent on Intel products to fill the fabs.
This would make the divisions more resilient should the other falter.
It may make for better resilience, but do you think our Defense Dept would let the only US fab go into foreign hands? That happened last century and they won't let it go further.
 

Vattila

Senior member
Oct 22, 2004
568
645
136
do you think our Defense Dept would let the only US fab go into foreign hands?
IBM has defence contracts that had to be catered for in the deal with GlobalFoundries. The latter is based in the USA, by the way. In general, a joint venture can be set up to operate from USA.
 

whm1974

Diamond Member
Jul 24, 2016
9,460
1,566
96
Thanks for all the replies and votes!

The consensus from the poll is "no", Intel will not go fabless.

However, I still think they will need to do something.
  • Spin off the manufacturing division into a fully owned subsidiary, which is run as an independent foundry, or
  • Create a joint venture with e.g. GlobalFoundries or Samsung, retaining majority ownership and control, or
  • Create a joint venture with a minority stake and a wafer supply agreement, or
  • Fully divest the fabs, and secure a wafer supply agreement as part of the deal.
The key, in my view, is to mitigate risk:
  1. Ensure that the design division can do diligent contingency planning, and opt to purchase wafers on the market, if the manufacturing division lags behind.
  2. Operate the manufacturing division as a foundry, catering to customer's needs, as to build a broad customer base, thus ensuring the scale needed to cover the cost of staying on the leading edge, while not being wholly dependent on Intel products to fill the fabs.
This would make the divisions more resilient should the other falter.
Actually Intel could easily manufacture other semiconductor devices besides CPUs and chipsets. In fact they are already doing that with NAND flash and Octane.
 

ehume

Golden Member
Nov 6, 2009
1,511
72
91
IBM has defence contracts that had to be catered for in the deal with GlobalFoundries. The latter is based in the USA, by the way. In general, a joint venture can be set up to operate from USA.
IBM did not operate the last US-based fab company. Intel does. Also, we have had a change of philosophies at the top.
 

bsp2020

Member
Dec 29, 2015
87
68
91
Thanks for all the replies and votes!

The consensus from the poll is "no", Intel will not go fabless.

However, I still think they will need to do something.
  • Spin off the manufacturing division into a fully owned subsidiary, which is run as an independent foundry, or
  • Create a joint venture with e.g. GlobalFoundries or Samsung, retaining majority ownership and control, or
  • Create a joint venture with a minority stake and a wafer supply agreement, or
  • Fully divest the fabs, and secure a wafer supply agreement as part of the deal.
The key, in my view, is to mitigate risk:
  1. Ensure that the design division can do diligent contingency planning, and opt to purchase wafers on the market, if the manufacturing division lags behind.
  2. Operate the manufacturing division as a foundry, catering to customer's needs, as to build a broad customer base, thus ensuring the scale needed to cover the cost of staying on the leading edge, while not being wholly dependent on Intel products to fill the fabs.
This would make the divisions more resilient should the other falter.
I'm sure that Intel wishes that they can :
1. spin off their FAB
2. fix their process technology
3. attract more customers
4. still retain control and develop process that will serve Intel First(TM)

Looking at that wish list, makes me feel like I'm looking at my son. When he wants to make a deal with his sister, it's usually when he has nothing to offer but still wants access to what his sister has...

I'm not sure whether spinning off their FAB is possible. Spin off makes sense when the main issue is money. AMD spun off their FAB when the money was tight and they (both AMD and Mubadala) thought bringing more money to the FAB would solve the problem. But the FAB had both money problem and technology problem. GF ended up licensing Samung 14nm tech, just in time for Ryzen. Without Samung technology, AMD and GF probably would not exist today. GF bought IBM FAB to get access to 7nm technology and beyond.

What Intel needs more than money is a solution to their process technology problem. They still have money to invest in technology so they should spend it to solve that problem. Spinning off FAB won't magically solve what ails Intel FAB.

I don't think anyone in their right mind would be willing to put up resources to take over Intel FAB and serve Intel as its main customer and let the tail wag the dog, especially when the main issue is not money. The main issue with Intel FAB is process technology and, at this point, Intel does not seem to have what it takes to solve that problem.
 
  • Like
Reactions: moinmoin

Vattila

Senior member
Oct 22, 2004
568
645
136
IBM did not operate the last US-based fab company. Intel does.
If foreign ownership is frowned upon by USA government, then Intel will simply have to find a solution with American ownership — should it be the case, as I posit, that their current integrated business model is too risky and unsustainable.
 

ehume

Golden Member
Nov 6, 2009
1,511
72
91
I'm sure that Intel wishes that they can :
1. spin off their FAB
2. fix their process technology
3. attract more customers
4. still retain control and develop process that will serve Intel First(TM)

Looking at that wish list, makes me feel like I'm looking at my son. When he wants to make a deal with his sister, it's usually when he has nothing to offer but still wants access to what his sister has...

I'm not sure whether spinning off their FAB is possible. Spin off makes sense when the main issue is money. AMD spun off their FAB when the money was tight and they (both AMD and Mubadala) thought bringing more money to the FAB would solve the problem. But the FAB had both money problem and technology problem. GF ended up licensing Samung 14nm tech, just in time for Ryzen. Without Samung technology, AMD and GF probably would not exist today. GF bought IBM FAB to get access to 7nm technology and beyond.

What Intel needs more than money is a solution to their process technology problem. They still have money to invest in technology so they should spend it to solve that problem. Spinning off FAB won't magically solve what ails Intel FAB.

I don't think anyone in their right mind would be willing to put up resources to take over Intel FAB and serve Intel as its main customer and let the tail wag the dog, especially when the main issue is not money. The main issue with Intel FAB is process technology and, at this point, Intel does not seem to have what it takes to solve that problem.
After all, IBM had to pay GF money. Did GF just take IBM's engineers and IP? Or did it take their fabs, too?
 

Vattila

Senior member
Oct 22, 2004
568
645
136
Did GF just take IBM's engineers and IP? Or did it take their fabs, too?
The acquisition included IBM's fabs in East Fishkill, NY and Essex Junction, VT. IBM and GlobalFoundries have joint R&D activities at SUNY Polytechnic in Albany, NY. And GlobalFoundries leading-edge Fab 8 is located in Saratoga County, NY with thousands of workers.

https://www.globalfoundries.com/news-events/press-releases/globalfoundries-completes-acquisition-of-ibm-microelectronics-business

For GlobalFoundries' presence in USA and around the world, see:

https://www.globalfoundries.com/about-us/contact-us/worldwide-locations
 
Last edited:
  • Like
Reactions: Cerb

ehume

Golden Member
Nov 6, 2009
1,511
72
91
The acquisition included IBM's fabs in East Fishkill, NY and Essex Junction, VT. IBM and GlobalFoundries have joint R&D activities at SUNY Polytechnic in Albany, NY. And GlobalFoundries leading-edge Fab 8 is located in Saratoga County, NY with thousands of workers.

https://www.globalfoundries.com/news-events/press-releases/globalfoundries-completes-acquisition-of-ibm-microelectronics-business

For GlobalFoundries' presence in USA and around the world, see:

https://www.globalfoundries.com/about-us/contact-us/worldwide-locations
Thanks for the info.
 

maddie

Diamond Member
Jul 18, 2010
3,390
2,338
136
I'm confused. These days when so many economies have been integrated, how do you determine ownership, especially a publicly traded company, from a nation state perspective?

Where registered, main shareholders, physical assets?
 

ehume

Golden Member
Nov 6, 2009
1,511
72
91
I'm confused. These days when so many economies have been integrated, how do you determine ownership, especially a publicly traded company, from a nation state perspective?

Where registered, main shareholders, physical assets?
All defined by statute. But I'll bet you won't want to see your fab dismantled and sent to China, as so many factories were.
 
  • Like
Reactions: maddie

bononos

Diamond Member
Aug 21, 2011
3,722
68
91
All defined by statute. But I'll bet you won't want to see your fab dismantled and sent to China, as so many factories were.
True I would expect Intel's foundry business be protected as a 'national security' asset.
 

LTC8K6

Lifer
Mar 10, 2004
28,523
1,573
126

bsp2020

Member
Dec 29, 2015
87
68
91
Why hasn't this company cut the losses and moved on to another fab or taken some other step? 10nm is years late, not months.
14nm++ seems to be as good or better than 10nm, why not use 14nm++?
I assume because it takes years to design a chip? When designing a chip, you plan years ahead. If the company started the design effort a couple of years ago, thinking that Intel would be manufacturing 10nm products in volume this year and they will be shipping their chip next year on a mature process, 2 years delay would put them in this position. Intel could has been telling the customer that they will resolve the issue soon until it was too late.

Not everyone has deep pockets like Intel, you know. It's not easy for someone to bite the bullets and take the loss and move on when you are half way into your project. For most people, it seems safer to wait for the problem to work itself out rather than scrap a project and start over half way through, especially if the big guy in the room keeps telling you that things will work out...
 

maddie

Diamond Member
Jul 18, 2010
3,390
2,338
136
Why hasn't this company cut the losses and moved on to another fab or taken some other step? 10nm is years late, not months.
14nm++ seems to be as good or better than 10nm, why not use 14nm++?
Following sounds familiar? Maybe they think like you.

https://forums.anandtech.com/threads/speculation-intel-will-become-fabless.2549870/#post-39479276
"Intel will be just fine, as will their fabs.

10nm will turn out to be fine, even though the risk of trying to do it their way ended up as a train derailment.
Train wrecks get cleaned up, the train gets re-railed, and the train slowly gets moving again.

7nm on EUV will be very nice for Intel fans.

Soon this bad patch will be in Intel's rear view mirror, getting smaller."
 

Stuka87

Diamond Member
Dec 10, 2010
5,376
1,166
136
And what is the name of this 20+ billion tech giant ? Any ideas ?
I would be willing to bet its Apple. Apple has already hinted that they were designing their own chips for desktop use. Other than them, there aren't any companies that come to mind.
 

scannall

Golden Member
Jan 1, 2012
1,776
1,227
136
I would be willing to bet its Apple. Apple has already hinted that they were designing their own chips for desktop use. Other than them, there aren't any companies that come to mind.
Apple is worth quite a bit more than $20 billion. ;-)
 

ASK THE COMMUNITY