[Motley Fool] Why Intel Corp. May Eventually Go Fab-less

[Tsavo]

It's about time they've seen the light.

I've been fabless my entire life. If it worked for me, it can work for Intel.

 

[Systems analyst]

Intel have signalled repeatedly that they are no longer a PC company; by 2020, I believe that they will have a fraction of the notebook market. Presumably, they have used cash-flow from the PC business to fund development of new generations of micro-code and production process nodes; without substantial profits from PCs, I cannot understand how they will fund new designs and processes. They seem to be transforming into a totally different company to the classical Intel, sans mobile and PCs.

 

[Sheep221]

Intel have signalled repeatedly that they are no longer a PC company; by 2020, I believe that they will have a fraction of the notebook market. Presumably, they have used cash-flow from the PC business to fund development of new generations of micro-code and production process nodes; without substantial profits from PCs, I cannot understand how they will fund new designs and processes. They seem to be transforming into a totally different company to the classical Intel, sans mobile and PCs.

They were never a PC company, they are electronics company

 

[VirtualLarry]

They were never a PC company, they are electronics company

But they've been the mfgs of the primary source of the heart and soul of the IBM-compatible PC since it existed.

 

[NTMBK]

But they've been the mfgs of the primary source of the heart and soul of the IBM-compatible PC since it existed.

And before that they were a memory manufacturer. Only the paranoid survive...

 

[CHADBOGA]

Are you aware that Intel fabs are more advanced than samsung, glofo and tsmc combined?

Sure.

But the gap is closing and the trend is a concern.

 

[Sheep221]

But they've been the mfgs of the primary source of the heart and soul of the IBM-compatible PC since it existed.

Yep, and they will still be, the fact that last year PC sales dropped few percent down encouraged almost everyone online to predict nearby apocalyptic future where all those billions computers on this planet become inferior niche technology and will be replaced by phones and tablets and if intel does not do anything, they will go bankrupt in a week!
As mentioned before, it won't happen, Intel is milking record revenues from selling rebadged generations of core i CPUs, only ones who are likely to stop buying PCs are our grandparents at best, but very much no one else, so no party is going on for these Intel and PC haters.

Oh I forgot to mention, ARM mobile SoCs are also designed on those evil and inferior bulky desktop PCs. Oh my god we are doomed!

 

[norseamd]

Can someone please tell me how the word combined has any real meaning here?

 

[norseamd]

Oh I forgot to mention, ARM mobile SoCs are also designed on those evil and inferior bulky desktop PCs.

You can probably design ARM CPUs and SOCs on computers made out of ARM processors also. Nevermind that POWER 8 should do anything and everything X86 does, but even better. Why buy a X86 computer when you can buy a POWER 8 computer?

 

[Sheep221]

Can someone please tell me how the word combined has any real meaning here?

It doesn't, it was only meant to describe that technological gap between other companies and Intel is much bigger than other semiconductor companies have between each other. Since they all are making much simplier chips than Intel, they can stack up and still be far less advanced than Intel is, that doesn't mean they are not important or their products are not important, they are, very however that's not what I meant either.

You can probably design ARM CPUs and SOCs on computers made out of ARM processors also. Nevermind that POWER 8 should do anything and everything X86 does, but even better. Why buy a X86 computer when you can buy a POWER 8 computer?

That's not what I was referring to, but ok, those are server CPUs, so such a workstation would be very likely similarly sized or larger than a x86 desktop PC. If it executes different instructions is not so important in this regard, all these forum threads are more focused on form factor rather than actual instruction sets. I do refer to x86 because it's most common.

 

[norseamd]

It doesn't, it was only meant to describe that technological gap between other companies and Intel is much bigger than other semiconductor companies have between each other. Since they all are making much simplier chips than Intel, they can stack up and still be far less advanced than Intel is, that doesn't mean they are not important or their products are not important, they are, very however that's not what I meant either.

Intel is currently notably more advanced, but the changes in process node technology, AKA Moores Law coming to an end, mean that the lead may effectively start to vanish even if they are first to every new process node, since it will take years for a new process node to come out, and eventually the fabs may be on the same node as Intel more of the time than Intel is actually ahead anymore. And honestly I wonder if it is even possible to go lower than 5nm to 7nm. I think new designs and technologies will start to matter far more than process node size.

 

[poofyhairguy]

If it executes different instructions is not so important in this regard, all these forum threads are more focused on form factor rather than actual instruction sets.

Yup, the ATOT world divides into nice product categories:

-Scary Servers

-Dinosaur Desktops

-Left Behind Laptops (aka DRs)

-Lean and Mean Laptops (Ultras)

-Modern and Magnificent Mobile Devices

-etc.

 

[Arachnotronic]

It will be interesting to see how long the timer runs before anyone catches up when Intel introduces Tunneling FETs or spintronics in 2019/2020 at 7nm.

You are a very creative guy Intel's 10nm is said to be in extremely poor shape right now, they'll be very lucky to hit their 2H 2017 timeline in any real quantities. Don't even think about Intel 7nm for a while, put this node out of your mind for now.

 

[Arachnotronic]

http://www.fool.com/investing/general/2016/05/02/why-intel-corp-may-eventually-go-fab-less.aspx

My guess would be the opposite, actually- Intel could spin off the CPU business, and go full on foundry.

Given how great Intel's foundry efforts have gone so far (you even correctly identified it as "kiss of death" for companies to try to port their IP to Intel process), this is a no go. After all, why do you think Stratix 10 from Altera is about a year behind schedule? Why do you think that Apple chose Samsung/TSMC for A9 instead of Intel?

If Intel doesn't fix the very serious problems with their process development methodology (and yes they are very serious, hearing that 10nm is a complete mess right now), then they are going to be "IBM'd" within the next 5-10 years. The market is going to be filled with 10nm TSMC/Samsung silicon in 2017 while I think Cannonlake-U/Y will be available in pretty much "trickle" quantities.

Mark Bohr can talk about Intel's superior density until he's blue in the face, but timing matters, too. This is the same crap we saw from IBM, always announcing some super amazing process technology/whiz-bang feature but never actually bringing these technologies to the point where they could be manufactured at high yields and reasonable cost.

The first shoe dropped with the 14nm node, and I think that 10nm will be just as bad, if not worse.

 

[witeken]

Given how great Intel's foundry efforts have gone so far (you even correctly identified it as "kiss of death" for companies to try to port their IP to Intel process), this is a no go. After all, why do you think Stratix 10 from Altera is about a year behind schedule? Why do you think that Apple chose Samsung/TSMC for A9 instead of Intel?


The first shoe dropped with the 14nm node, and I think that 10nm will be just as bad, if not worse.

The porting you have to do just once, so who cares if it has benefit in the long term (if Intel maintains process lead).

And we'll see about 10nm. That should mean that TSMC/SS 7nm should be just as hard.

 

[NTMBK]

Given how great Intel's foundry efforts have gone so far (you even correctly identified it as "kiss of death" for companies to try to port their IP to Intel process), this is a no go. After all, why do you think Stratix 10 from Altera is about a year behind schedule? Why do you think that Apple chose Samsung/TSMC for A9 instead of Intel?

Right now, certainly. But I suspect that this is a symptom of Intel's underlying mentality, the same one that destroyed their chances in mobile- the foundry work is a "side project". It's meant to bring in some extra revenue on the side, but it's never going to take priority over Intel's own Core and Xeon development. If that x86 CPU business were just another external customer, perhaps the foundry business would be taken a little more seriously.

Sure, it's a long shot that they could ever make it work. But it seems like a fundamentally more appealing prospect than being a fabless x86 manufacturer tied to a shrinking PC market, with no manufacturing advantage moat to protect your server castle.