Speculation: Intel will become fabless

Page 15 - 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: 24 13.4%
  • No, manufacturing is integral to Intel, and they will continue to invest to stay competitive.

    Votes: 155 86.6%

  • Total voters
    179

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
Can they do contingency planning on the cheap without affecting beloved margins? They are already taking a painful cut to their gross margins as they ramp up production for Tiger Lake and upcoming products on their expensive low-yielding 10nm process.

They already lost ton of margins because they lost the competitive edge. 7% off from their peak when they were arguing 2% differences at one point.

If they are competitive again, they can reduce such "margin pressure" for the product team while keeping the 56% margin they are at now(net margin is much lower, 30% now, 21% before 2018). I think long term it would be much better.
 
  • Like
Reactions: Tlh97 and Vattila

ThatBuzzkiller

Golden Member
Nov 14, 2014
1,120
260
136
Intel declining in a high-tech manufacturing sector could be a sign of what's to come for all of silicon valley. It could soon become the next rust belt with a "worn down" political demographic like we saw in the last election ...

Much of it's talent in the next decade will become 'hacks' working menial jobs because foreign competition destroyed all of their careers ...
 

A///

Diamond Member
Feb 24, 2017
4,352
3,154
136
Intel declining in a high-tech manufacturing sector could be a sign of what's to come for all of silicon valley. It could soon become the next rust belt with a "worn down" political demographic like we saw in the last election ...

Much of it's talent in the next decade will become 'hacks' working menial jobs because foreign competition destroyed all of their careers ...
They'll just beg the government for 500B-1T like they did in the 1980s.
 

Vattila

Senior member
Oct 22, 2004
799
1,351
136
1616607561706.png

Intel’s New IDM 2.0 Strategy: $20b for Two Fabs, Meteor Lake 7nm Tiles, New Foundry Services, IBM Collaboration, Return of IDF (anandtech.com)

Wow! Gelsinger is pretty bold in his enthusiastic and aggressive move to double down on industry domination and the risk/money he is willing to take/spend to achieve it. It is pretty scary — for AMD and Intel shareholders alike.

In the long term, should it succeed, and enable Intel to retain its dominant position and market share, then it will clearly threaten AMD's success and growth path. If it fails, then Intel will be in even bigger problems than today, having invested even more in a currently failing IDM model.

In the short term, there is nothing new in the presented plan that will adversely affect AMD's road map for the next two years, while in the face of increasing competition, Intel shareholders may see declining gross margin and revenue ($72B projected for this year, while they have been above that in their trailing twelve months revenue for the last four quarters; see Macrotrends), while spending goes up dramatically (from $14B last year to $19-20B projected this year). Which means less dollars spent on propping up the share price with stock buybacks and dividend increases.

I'm a little bit baffled about the Intel Foundry Services business. As an independently run business, it looks like a serious attempt at a foundry play. But I guess it is just a safe-guard to sell off excess capacity, to justify expansion, and as an insurance, in case they cannot keep their fabs fully utilised with their own product success. If I understand correctly, the fabs will still be in the possession of the main company, which is doubling down on the IDM model. In short, I suspect there is not much foundry ambition within Gelsinger. If he was serious about it, he would have spun off the fabs properly and partnered with an existing foundry to create a truly independent entity. I get the impression he simply wants back to Intel glory, with market share sufficient to fill their fabs with Intel product.

As pointed out by many, with this process leadership plan from Gelsinger, not only does TSMC have no incentive to be helpful to Intel — it has all the incentives to be helpful to AMD.
 
Last edited:

blckgrffn

Diamond Member
May 1, 2003
9,111
3,029
136
www.teamjuchems.com
"Which means less dollars spent on propping up the share price with stock buyback and dividend increases."

I read this into it too. Heck yes. Have enough faith in your own business to invest in it rather than stock buy back garbage.

I can't express adequately how important I feel like this change of course is.

Even if if "fails" ultimately I think it is the better play for *everyone* in the long term.
 

Mopetar

Diamond Member
Jan 31, 2011
7,797
5,899
136
I'm not an optometrist, but even I can tell that Gelsinger has a vision that can see beyond the next quarter.
 

jpiniero

Lifer
Oct 1, 2010
14,510
5,159
136
Isn't the US Government paying for some/all of that 20 billion? Bringing back the Foundry Services was presumably part of the deal since it would give defense (including IBM Power?) another option than the TSMC deal for products that need to be manufactured domestically.

Doesn't mean it will work..
 

gdansk

Golden Member
Feb 8, 2011
1,987
2,356
136
Isn't the US Government paying for some/all of that 20 billion?
Where did you get that idea? According to the Anandtech article Intel is financing it:
Intel will be announcing a $20 billion (USD) investment in two new manufacturing facilities (fabs) in Arizona, set to come online for production in 2024.

And the NYT corroborate that it is Intel doing the investment with incentives from local and federal authorities. This is the same structure Intel and Samsung have used in the past for previous factories. Samsung and TSMC are doing the same for their upcoming Austin/Arizona fabs. They usually come in the form of tax credits for permanent job creation and are not part of the investment figure of $20 billion.
 
Last edited:
  • Like
Reactions: Tlh97 and Vattila

DrMrLordX

Lifer
Apr 27, 2000
21,582
10,785
136
I'm a little bit baffled about the Intel Foundry Services business.

I am too.

Why build out fabs if you don't have a good process plan? 10SFE is very late to market and 7nm is completely MIA. What nodes will they produce at these fabs? Who will be their customers? Nobody bit on their offer to provide foundry services in the past - at least nobody significant, and various iterations of 14nm were on the table at that time. Intel mostly wound up buying other companies and moving those products to Intel fabs. I don't see the play here unless Intel is planning on sticking a fork in their IDM model, spinning off all their fabs, and using this as a toe-in-the-water approach to see if they can get established in the foundry sector.
 
  • Like
Reactions: Tlh97 and Vattila

gdansk

Golden Member
Feb 8, 2011
1,987
2,356
136
I'd wager 1 INTC stock it's going to be focused on 14nm to start. They built out more 14nm lines than they had originally intended. And soon Intel's main products will be moving over to 10nm, external fabs and their own 7nm. There will not be enough machines to retrofit those lines with the required EUV equipment. If you can't retrofit it and will be using it less yourself, what do you do? Foundry! Plus it would be a good process for embedded customers, much of which is still on 40 and 28nm.

It is a key part of preparing Intel fabs for spin off, probably under the next CEO. It would need external customers on their older nodes, like Samsung and TSMC have, to ever have a shot as a spin off.
 
  • Like
Reactions: Tlh97 and Vattila

Doug S

Platinum Member
Feb 8, 2020
2,203
3,405
136
Isn't the US Government paying for some/all of that 20 billion? Bringing back the Foundry Services was presumably part of the deal since it would give defense (including IBM Power?) another option than the TSMC deal for products that need to be manufactured domestically.

Doesn't mean it will work..

Intel said all $20 billion committed will be their own money, and doesn't count any additional incentives they may receive.

Given the efforts in both the US and EU for more self sufficiency in this area, no doubt there will be billions coming their way and that $20 billion will grow. Intel already has fabs in Ireland, so the EU side of this might expand there or might go elsewhere there is / has been high end fab talent like Germany.
 
  • Like
Reactions: Tlh97 and Vattila

jpiniero

Lifer
Oct 1, 2010
14,510
5,159
136
I'd wager 1 INTC stock it's going to be focused on 14nm to start. They built out more 14nm lines than they had originally intended. And soon Intel's main products will be moving over to 10nm, external fabs and their own 7nm. There will not be enough machines to retrofit those lines with the required EUV equipment. If you can't retrofit it and will be using it less yourself, what do you do? Foundry! Plus it would be a good process for embedded customers, much of which is still on 40 and 28nm.

That would make a lot of sense but that's not what Intel has done in the past. Intel usually reuses space with new nodes.
 

LightningZ71

Golden Member
Mar 10, 2017
1,627
1,898
136
WRT intel's domination intentions being a problem for AMD; AMD's biggest problem continues to be their own forced decision to spin off their fab business. As long as the likes of Intel, Nvidia, and Apple have massively deep pockets to bid up the cost of leading edge nodes at TSMC and likely Samsung as well, they will always have to fight for and overbid for foundry space at those fabs. They will continually be at a perpetual cost disadvantage or capacity constrained on competitive nodes.

The current situation is all the proof you need. We're over six months into volume commercial production on N5, and even longer than TSMC began to offer N6, and, so far, we haven't seen AMD so much as sniff either of those nodes. It's going to be at least another quarter before AMD even starts making noise about a product on either one, and it's going to be like a leaking faucet trying to fill an olympic swimming pool when they start. It's not going to matter that Intel is having node difficulties as AMD won't be able to feasibly produce enough product on any better node to make any real market impact.

Intel only has to not screw up in a major way to maintain the substantial market dominance that they currently enjoy, even if AMD does even better on the R&D side than we all hope in our wildest dreams.
 
  • Haha
Reactions: scineram

itsmydamnation

Platinum Member
Feb 6, 2011
2,744
3,078
136
WRT intel's domination intentions being a problem for AMD; AMD's biggest problem continues to be their own forced decision to spin off their fab business. As long as the likes of Intel, Nvidia, and Apple have massively deep pockets to bid up the cost of leading edge nodes at TSMC and likely Samsung as well, they will always have to fight for and overbid for foundry space at those fabs. They will continually be at a perpetual cost disadvantage or capacity constrained on competitive nodes.
thats like completely wrong......
 

gdansk

Golden Member
Feb 8, 2011
1,987
2,356
136
WRT intel's domination intentions being a problem for AMD; AMD's biggest problem continues to be their own forced decision to spin off their fab business. As long as the likes of Intel, Nvidia, and Apple have massively deep pockets to bid up the cost of leading edge nodes at TSMC and likely Samsung as well, they will always have to fight for and overbid for foundry space at those fabs. They will continually be at a perpetual cost disadvantage or capacity constrained on competitive nodes.

The current situation is all the proof you need. We're over six months into volume commercial production on N5, and even longer than TSMC began to offer N6, and, so far, we haven't seen AMD so much as sniff either of those nodes. It's going to be at least another quarter before AMD even starts making noise about a product on either one, and it's going to be like a leaking faucet trying to fill an olympic swimming pool when they start. It's not going to matter that Intel is having node difficulties as AMD won't be able to feasibly produce enough product on any better node to make any real market impact.

Intel only has to not screw up in a major way to maintain the substantial market dominance that they currently enjoy, even if AMD does even better on the R&D side than we all hope in our wildest dreams.
AMD may not have the volume & margin required to finance construction of fabrication facilities. I think this was clear back when Global Foundries cancelled their 7nm plans (despite collaboration with Samsung and IBM). Capacity may be holding them back right now* but that would be solved more easily and with more flexibility by having a product mix at multiple different foundries. Obviously it is too late to help the current situation but the impact of future shortages could be reduced this way.

* I'm not even sure that's crux of the problem. DigiTimes has reported very long lead times in all sorts of ancillary industries such as packaging. No one predicted such an unusual uptick in demand.
 

LightningZ71

Golden Member
Mar 10, 2017
1,627
1,898
136
Even if AMD managed to claim 50% of all of TSMC'S two most advanced nodes production capacity, that wouldn't amount to even twenty percent of Intel's volume on 14 and 10nm. In gaining that amount of production, they would have gone through a bidding war against two or more giant companies that have enough money in the bank to buy AMD several times over. TSMC is NOT a charity. They aren't going to give AMD anything.

Until AMD has enough money to build their own fab, or, perhaps, enter into a constructive partnership with TSMC or Samsung to build a fab that TSMC or Samsung operate on their behalf, they will always be nearly irrelevant, representing a tantalizing amount of competition that's just enough to keep the regulators off their backs about having a monopoly.

AMD is currently executing well with a solid business plan and aggressive, but manageable goals. They are not going to dictate the market. They have only recently been given enough respect in the laptop space to land a few mid-tier and near premium design wins, but the vast majority of their CPUs wind up in margin basement units with crummy screens and low end video cards. They are barely a spec on a housefly's rear end in the corporate market. If they want more volume, they are going to have to pay through the nose for it. It's likely not even worth it to produce more than they currently do given their situation.
 
  • Haha
Reactions: scineram

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
In short, I suspect there is not much foundry ambition within Gelsinger. If he was serious about it, he would have spun off the fabs properly and partnered with an existing foundry to create a truly independent entity. I get the impression he simply wants back to Intel glory, with market share sufficient to fill their fabs with Intel product.

We don't know the details so it's too early to say that.

They didn't fail in the past because their process sucked, but lacked the necessary details that serve customers as a Foundry. Maybe the customers had to use Intel proprietary tools and spend extra time and resources to learn it. Maybe the recipes available weren't fully available to the customers, and it would have been hard to make a full SoC with an Intel process. After all, their main product line isn't an SoC.

Gelsinger is not only a long time Inteller but a star in the company and quickly rose to CTO ranks until he got sacked during the Larrabbee days by then-CEO Paul Otellini.

He was part of the team that was responsible for the 386 and basically saved the company since iAXP432 tanked, and became the lead architect for 486. Gelsinger was also the part of Intel that recognized the "power wall" and the endless chase of frequency will result in a disaster.

If anyone can work out the details, it's him.
 

DrMrLordX

Lifer
Apr 27, 2000
21,582
10,785
136
I'd wager 1 INTC stock it's going to be focused on 14nm to start. They built out more 14nm lines than they had originally intended. And soon Intel's main products will be moving over to 10nm, external fabs and their own 7nm. There will not be enough machines to retrofit those lines with the required EUV equipment. If you can't retrofit it and will be using it less yourself, what do you do? Foundry! Plus it would be a good process for embedded customers, much of which is still on 40 and 28nm.

It is a key part of preparing Intel fabs for spin off, probably under the next CEO. It would need external customers on their older nodes, like Samsung and TSMC have, to ever have a shot as a spin off.

The only thing that sticks out to me is that Intel already offered nodes like 22nm and 14nm to foundry customers, and that resulted in very few takers. They had a deal with Rockchip that went mostly nowhere. Yeah there's a thirst for fab capacity, but remember that Intel's existing IP is almost exclusively tuned to meet the needs of their own designs (and vice versa). Embedded customers on nodes as old as 40nm have had the chance to take Intel 22nm wafers for years (at least in theory) and it doesn't seem to be happening much, if at all.
 

Stuka87

Diamond Member
Dec 10, 2010
6,240
2,559
136
Even if AMD managed to claim 50% of all of TSMC'S two most advanced nodes production capacity, that wouldn't amount to even twenty percent of Intel's volume on 14 and 10nm.

Where are you getting this data? You are basically saying that Intel is capable of making over twice as many chips as TSMC is?

As of today, we know this isn't true, so why make this claim?

Now if Intel's plans for new fabs actually works out, then this may change. But as Intel has zero influence in the mobile device world, I am unsure how much it will help them. As the x86 (desktop/laptop/server) market is only so big.
 
  • Like
Reactions: Tlh97

LightningZ71

Golden Member
Mar 10, 2017
1,627
1,898
136
TSMC, as of the end of 2019, represented just under 13% of the world's total wafer production capacity. This includes ALL of their nodes, including a bunch of older, non-competitive nodes. Their N7 and N5 capacity is much less than this total. That's not even up for debate, those are their published figures.

Intel has 11 300mm fabs, all of which can produce 14nm or better wafers and two more in the process of being built. While they still make things on older nodes, the vast majority of their production is on what is considered a competitive leading edge node (I'm including 14nm in this because, as we see, Rocket Lake is still competitive with other market leaders).

My comments are SPECIFICALLY RELATED to LEADING EDGE NODES. Intel's capacity there dwarfs TSMC, who is currently AMD's only choice for a competitive product. Samsung's 8nm node is not quite there, and Nvidia has that heavily engaged.

When we're talking about AMD vs. Intel, it is entirely the x86 processor (and soon to be GPU) market. Intel produces significantly more x86 processors than AMD and AMD doesn't have anywhere to go to get more capacity on a competitive node. Intel has deep, deep pockets, and has already announced that they are purchasing capacity from TSMC. Apple is going to be needing more processor capacity as they replace Intel sourced products with their own processors. Apple is a big purchaser of TSMC capacity, and that's only going to grow as they purchase more wafers to satisfy their need for larger processors for their higher end products.

Where, exactly, is this Capacity Fairy that's going to magically solve AMD's capacity limitations?
 

gdansk

Golden Member
Feb 8, 2011
1,987
2,356
136
The only thing that sticks out to me is that Intel already offered nodes like 22nm and 14nm to foundry customers, and that resulted in very few takers. They had a deal with Rockchip that went mostly nowhere. Yeah there's a thirst for fab capacity, but remember that Intel's existing IP is almost exclusively tuned to meet the needs of their own designs (and vice versa). Embedded customers on nodes as old as 40nm have had the chance to take Intel 22nm wafers for years (at least in theory) and it doesn't seem to be happening much, if at all.
New CEO seems to think initial attempt was flawed and I tend to agree. Intel had only accepted designs that included their IP. Then Intel's capacity shortage removed all reason to make external designs. This time it seems they'll make anything with or without Intel IP and won't have another 14nm capacity shortage (thanks TSMC/AMD).
 
  • Like
Reactions: Vattila

Stuka87

Diamond Member
Dec 10, 2010
6,240
2,559
136
TSMC, as of the end of 2019, represented just under 13% of the world's total wafer production capacity. This includes ALL of their nodes, including a bunch of older, non-competitive nodes. Their N7 and N5 capacity is much less than this total. That's not even up for debate, those are their published figures.

Intel has 11 300mm fabs, all of which can produce 14nm or better wafers and two more in the process of being built. While they still make things on older nodes, the vast majority of their production is on what is considered a competitive leading edge node (I'm including 14nm in this because, as we see, Rocket Lake is still competitive with other market leaders).

My comments are SPECIFICALLY RELATED to LEADING EDGE NODES. Intel's capacity there dwarfs TSMC, who is currently AMD's only choice for a competitive product. Samsung's 8nm node is not quite there, and Nvidia has that heavily engaged.

When we're talking about AMD vs. Intel, it is entirely the x86 processor (and soon to be GPU) market. Intel produces significantly more x86 processors than AMD and AMD doesn't have anywhere to go to get more capacity on a competitive node. Intel has deep, deep pockets, and has already announced that they are purchasing capacity from TSMC. Apple is going to be needing more processor capacity as they replace Intel sourced products with their own processors. Apple is a big purchaser of TSMC capacity, and that's only going to grow as they purchase more wafers to satisfy their need for larger processors for their higher end products.

Where, exactly, is this Capacity Fairy that's going to magically solve AMD's capacity limitations?

My question is where are you getting your data from?

Articles such as this tell a very different story from what you are saying: https://www.electronicsweekly.com/news/business/capacity-leaders-wafer-size-2021-02/

Here is a chart from that article, which is from last month:
11c94e25-4b44-4435-99c2-63625e68fc83.png


EDIT: Fixed image so that it displays
 
Last edited:

jpiniero

Lifer
Oct 1, 2010
14,510
5,159
136
  • Like
Reactions: beginner99

LikeLinus

Lifer
Jul 25, 2001
11,518
670
126
My question is where are you getting your data from?

Articles such as this tell a very different story from what you are saying: https://www.electronicsweekly.com/news/business/capacity-leaders-wafer-size-2021-02/

Here is a chart from that article, which is from last month:
My question is where are you getting your data from?

Articles such as this tell a very different story from what you are saying: https://www.electronicsweekly.com/news/business/capacity-leaders-wafer-size-2021-02/

Here is a chart from that article, which is from last month:
Curiosity, is the confusion because he is talking about wafers for CPU production and not memory? Your article says:

"It is not surprising that the 300mm ranking includes only DRAM and NAND flash memory suppliers like Samsung, Micron, SK Hynix, and Kioxia/WD; four of the world’s largest pure-play foundries TSMC, GlobalFoundries, UMC, and Powerchip (including Nexchip); and Intel, the industry’s biggest manufacturer of microprocessors.
 

Vattila

Senior member
Oct 22, 2004
799
1,351
136
WRT intel's domination intentions being a problem for AMD; AMD's biggest problem continues to be [manufacturing capacity and cost]

Let us not derail this thread into a discussion about AMD's supposed capacity problems. AMD is executing well beyond their financial growth plan of 20% CAGR, with decent and increasing gross margins, while reporting good visibility of wafer supply throughout 2021, sufficient to support the 37% revenue growth forecasted this year by CEO Lisa Su (likely with some headroom). If you are interested in more on my take on this, see my post here.

As for your view that an IDM model is superior to a fab-less business model, and that AMD should have its own fabs, my rationale for this whole thread is that the IDM model is not sustainable. I actually think Intel will have to spin off their fabs at some point. A single vendor will not have enough product volume to fund the R&D and capex needed to stay at the increasingly costly cutting edge of chip manufacture. Perhaps that progress will stagnate and leading-edge costs will come down at some point, but there is no sign of that for the foreseeable future.

Intel has managed to fund leading-edge fabs because they have dominated the industry and generated sufficient operating profits at stellar margins. But, as pointed out, fab cost is increasing relentlessly, while at the same time Intel's industry domination is facing serious threat from many directions; competition from AMD in the x86 space, increasing use of ARM in PC and server, increasing adoption of RISC-V in all fields, transition to GPGPU and accelerators for AI/ML and high-performance compute, etc. Intel is already seeing falling gross margins and stagnating revenue (notably, Pat Gelsinger forecasted a slight decrease for 2021). Gelsinger is aiming to reverse this trend, return Intel to glory, and generate the revenue growth and margins needed to not only keep up, but to again take the industry lead in chip manufacture. I think he might be too ambitious.

While I am sure Gelsinger wants the Intel Foundry Services business to succeed, it seems he intends to fund the fabs by Intel product success — not by scaling IFS to the level needed to be self-sufficient.
 
Last edited: