You misunderstand how the computing industry works. OEM customers don't choose fabrication nodes, they choose chips like CPUs. For example, building a laptop based on the Raptor Lake CPU. The fact that Raptor Lake is made on Intel 7 process is irrelevant in their decision. Price, performance, power consumption, quality and reliability, features, issues like that, drive their decisions. Fab processes are merely means to an end outside of their control to companies like Dell and Lenovo.
Forget the White House Sideshow. Intel Must Decide What It Wants to Be.
- Thread starter Fred Chen
- Start date
I don't think that's what it says
There is a pretty good discussion happing here between the lines on some of the factors impacting the economics of foundry.
I think people who get deep into this stuff can start to understand why IDM model is so broken... because at advanced nodes you need a lot of volume to make a foundry profitable, and that volume needs many customers and strong external relationships with those customers. If Intel is truly able to treat it's own design group as a completely independent external customer with no special preferences, and customers truly believe that to be the case maybe... but that's a lot to ask and in practice is probably impossible.
But you can also read about all the things Intel has done in the past that have not been exactly customer friendly so you already have a big wall to climb in terms of trust.
I think people who get deep into this stuff can start to understand why IDM model is so broken... because at advanced nodes you need a lot of volume to make a foundry profitable, and that volume needs many customers and strong external relationships with those customers. If Intel is truly able to treat it's own design group as a completely independent external customer with no special preferences, and customers truly believe that to be the case maybe... but that's a lot to ask and in practice is probably impossible.
But you can also read about all the things Intel has done in the past that have not been exactly customer friendly so you already have a big wall to climb in terms of trust.
Artificer60
Well-known member
And yet to succeed at foundry on the leading edge, you need that customer zero to work with while the node is still in development. I would argue that is Apple for TSMC. TSMC needs that feed back from the customer to finish developing the PDK on real world designs, not just SRAM. It only makes sense for that customer to be Intel products for Intel foundry.
I believe Intel understands what they need to do to be a successful foundry this time and have brought in people from the outside to help. But knowing what to do isn't the same thing as knowing how to be a foundry and that is going to take time. If Intel products is separated from the foundry, they no longer have incentive to invest the effort in helping Intel foundry get better and Intel Foundry loses that customer zero.
You can argue that this could be enforced with contracts at the time of the split, but I saw how that worked out with AMD/Global Foundries. My personal opinion is that keeping them together until Intel foundry can stand on their own feet is a much better approach. Zinsner indicated that at least for the next few years Intel foundry only needs a few billion dollars more a year from external customers to reach break even. Given time I believe that Intel foundry can be a successful leading edge foundry that continues to push the boundaries at the leading edge. But they need that customer zero to get there.
hist78
Well-known member
Yes, that’s the ideal approach.
But as we’ve seen with many companies, including Intel, once they begin making decent profits and business appears stable, or as you said, “until Intel Foundry can stand on its own feet” they tend to forget the fundamental changes that must be made. Instead, the focus shifts to debates over dividends, share buybacks, executive bonuses, employee stock options, salary increases, or compensation packages.
As a result, the necessary changes become inconvenient and are left unmentioned.
To save Intel, a split is essential to force a change, and no time can be wasted. The longer Intel waits, the smaller the chance that either or both divisions, the Product division or the Foundry division, will survive. In the worst case, Intel could face a fire sale that would be bad to everyone.
that's a little too harsh. Intel still holds x86 and all its patents. That's gotta be worth something
But to who? Who -- apart from AMD -- could make use of them? The barrier to a successful x86 CPU market entry is enormous, as all the other companies who have tried and failed could testify -- if they still existed...
hist78
Well-known member
The IDM model may have been great 20 or 30 years ago, but the world today is very different. Intel and many of its past leaders have been so accustomed to the old, dominant Intel that they couldn’t find a way to move the company forward in the 21st century. Worse, some of them were so indoctrinated by the IDM model that they failed to recognize Intel was already on a dangerous path. At times, their behavior and decisions were not only odd but also damaging.
For example, in 2021, then-CEO Pat Gelsinger launched a series of video campaigns attacking Apple and Mac computers, just as Apple was transitioning its Mac lineup from Intel CPUs to its own M series processors. More importantly, Gelsinger had already announced Intel’s IDM 2.0 strategy and was actively seeking customers for Intel Foundry. Why would Intel and its CEO willingly attack a long-time customer and potential foundry client?
Even stranger, the supposed weaknesses of Apple Macs highlighted in those Intel videos were actually design choices made by the operating system developers or hardware manufacturers, not the weakness of CPUs themselves. Because for many years, both Windows PCs and Apple Macs relied on Intel processors. If anyone were entitled to make such attacks, it would have been Microsoft, Dell, HP, Lenovo, Asus, or Acer, not Intel.
The most concerning part is that within Intel, across the board of directors and senior executives, no one stopped Gelsinger from pursuing such childish and damaging behavior.
Intel had removed those videos from Intel official YouTube channel. But you can still see partial clips from other YouTube channels.
siliconbruh999
Well-known member
I don't get it why are you against IDM Model like I have said choices and business decision matters not IDM.
The only part I agree with you is the part of attacking apple that was a bad thing.
On a sidenote in Apple's they have the most flexibility in designing chips Intel/AMD the least they don't get the praise for making everything compatible for 40+ years.
What happens when you put pressure on Apples you can't do this or that that will lead to design choices increase in TTM.
It's pretty easy to design a CPU that you only have to be able to run your software for you targeted audience with limited combination of HW.
They were able to pull this stuff cause the control their SW Ecosystem.
The only part I agree with you is the part of attacking apple that was a bad thing.
On a sidenote in Apple's they have the most flexibility in designing chips Intel/AMD the least they don't get the praise for making everything compatible for 40+ years.
What happens when you put pressure on Apples you can't do this or that that will lead to design choices increase in TTM.
It's pretty easy to design a CPU that you only have to be able to run your software for you targeted audience with limited combination of HW.
They were able to pull this stuff cause the control their SW Ecosystem.
The return on invested capital kind of shows what happens when you go from at-sufficent-scale to sub-scale in the fab/foundry business at the leading edge. The biggest issue that Intel really has it that is has to do two things with any foundry business - 1) remain at-scale at the leading edge, plus 2) figure out how to sell capacity in old, depreciated fabs rather than retire them.
Last edited:
Hopefully everyone at Intel who thought it was a good idea to attack Apple are gone. Apple could easily move the M5 chip to Intel Foundry 14A and I hope they do.
If fabless customers had (3) leading edge pure-play fabs to choose from I would agree with you on the IDM thing but they do not. There is one pure-play foundry with 90%+ market share at 3nm and 2nm and two IDM foundries who are struggling. One is in South Korea which has a very unstable Northern neighbor and the other is Intel who's innovations we have been relying on for the past 50 years. It is an easy choice really.
It's an easy choice when one of your suppliers costs less, delivers better performance, and is easier to work with. It's especially easy if the competitions only real selling point is that they might be a better choice in the case of WW3. People will talk about wanting options, but they will always go with the option that gives the best combination of cost, performance, and service. If Intel can't compete on those parameters I don't think they can compete period.
siliconbruh999
Well-known member
I don't think performance will be a big issues with Intel as Cost/Service would be cause. Intel doesn't have much IP and not much experience for service as for cost America is expensive to do this stuff and they don't have enough volume to offset the capital cost.
hist78
Well-known member
Isn’t it clear that the industry needs three strong foundries to ensure healthy competition? Yet the narrative is often stretched to suggest that the industry really needs is
If Intel cannot split itself while it is weak, what makes us believe it will do so once it regains strength? How can Intel’s product competitors trust that they are not simply feeding their own rival, the Intel IDM, today and in the future?
Are Intel’s competitors truly so naïve as to believe that once Intel becomes stronger and its business stabilizes, it will willingly split itself into two independent companies, Intel Products and Intel Foundry? And who could guarantee that outcome?
Last edited:
Samsung is an IDM.
No matter how old the manufacturing equipment is, it would be good if it could be sold as a mature process.
Most of the semiconductors in the world are mostly mature process products.
Maybe there are people who want to participate in the X86 ecosystem.
For Intel and AMD, it may become a future rival in some fields, but it may be very good in terms of expanding and maintaining the ecosystem.
I agree with you. Intel, like many other companies before it, has become a victim of its own success. What worked, worked so well, it could never be questioned. When it quit working, there was simply no appetite to even entertain the changes that had to be done. Worse I fear, is that the company culture within Intel is possibly incapable of making such a radical turn as to become a service company (Foundry).
I would bet good money that you are correct; however, I doubt that we will find out. It is my current opinion that dumping money into the same machine .... using the same cogs and mechanics as it had before, will simply suck up that money like water poured on the desert sand.
I also think that the Intel foundry has become the boogie man inside Intel (someone correct me if I am wrong). It seems like the losses are largely blamed on the foundry results.
Where else should the blame be placed? I am likely not qualified to say. My guess is that there were simply too many risky bets placed on too many endeavors that didn't play out favorably. I kind of see this in the foundry and in the processor architecture. It's not new either. I still recall the P4, Rambus and Itanium days. Like I said. Risky.
Still, Intel held the world in its hands for decades. As spectacularly as if failed (P4), it often recovered with a vengeance (Core 2). In the days of 14nm (and before) Intel dominated the lithography world sometimes holding a 2 node shrink advantage over all others (IIRC). It's almost impossible to imagine that they WOULDN'T keep holding to the recipe that kept them on top for decades.
IBM comes to mind. Anyone buy an IBM PC lately
. Perhaps this is a good example of a company utterly missing the boat, burning to the ground, and rising from the ashes. Surely IBM will be taught to perspective MBA students for decades to come. Intel may also become a subject for these students. (Off topic) I am very sure Tesla will as well .