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

[NTMBK]

At this rate, Intel may eventually find that there is no really differentiation associated with using its own manufacturing technology relative to TSMC's. Or, perhaps in a "doomsday" scenario, Intel might find that foundry technology has become better than its own.

In that case, it's not clear that it would make sense for Intel to continue soldiering on with its own manufacturing technology.

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.

 

[ShintaiDK]

That may be the single dumbest article yet from that guy

Intel still sits on ~65% of all global MPU revenue. The only revenue that matters. And why pay TSMC a huge chunk of your margins when you dont have to.

And with Apple in free fall, who is going to pay for the fabless node progress.

 

[VirtualLarry]

That may be the single dumbest article yet from that guy

Yep. Who was it that said "Only Real Men have FABs"? AMD's original CEO?

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

I agree. Intel's product designs aren't really awe-inspiring any more. More of the same old, same old, plus up to five percent. But they still have a foundry lead. How long that will last, is anyone's guess.

 

[ShintaiDK]

Yep. Who was it that said "Only Real Men have FABs"? AMD's original CEO?

Yep. Jerry Sanders (The third).

 

[witeken]

No chance, ever. Worst idea from Ashraf.

If Intel goes fabless, who will its (the Intel fab) customers be? Where will they get the volume and revenues that Intel gets? How will they stay on the leading edge. Spinning the fabs off is the best recipe for disaster.

Don't forget that most of TSMC's volume comes from old process nodes.

Intel doesn't need to spin off. If there is money to be earned in being a foundry, they can invest in that within Intel, although BK didn't talk about that so that's I guess another one of his failed attemps. (Apple would've been the best candidate.)

The whole point about being an IDM is that is has huge advantages. No foundry tax (the money from Intel and the fab would be split, which leads to huge money flow disadvantage), process/architecture integration and optimization, TTM improvement. (Only problem is maybe that Intel doesn't want to go into the 5x% gross margin so they can't afford to go to newer processes until they're mature enough.)

Edit: Seems others agree, wrote this comment before reading the other ones.

Edit: Also don't forget Intel's data center effort with silicon photonics, Omni Path, etc. Intel is at heart a hardware company, so spinning off their in-house manufacturing doesn't make any sense. Unless you have financial problems. But who is still the world's biggest semiconductor company?

 

[cytg111]

I can see the issue, where is the volume going to come from to sustain the R&D beast? Would hate for Intel to go fabless, it worked out so well for AMD as we all know.
Couldnt Intel fab ARM cores on the previous gen node or node before that? Chew out 22 and 32nm ARM socs left and right?

 

[NTMBK]

I can see the issue, where is the volume going to come from to sustain the R&D beast? Would hate for Intel to go fabless, it worked out so well for AMD as we all know.
Couldnt Intel fab ARM cores on the previous gen node or node before that? Chew out 22 and 32nm ARM socs left and right?

Intel processes are too expensive for that. There's a reason why SoFIA was made at TSMC.

 

[imported_ats]

That may be the single dumbest article yet from that guy

The other thing to realize is that Intel literally has 2-3 complete fabs on the leading edge of any given time. TSMC generally doesn't have even a single fab on the leading edge. Intel is currently processing more 14nm logic wafers than all other companies combined are processing 16/14nm wafers. The vast majority of TSMC revenues and profit are on trailing edge processes.

Intel still sits on ~65% of all global MPU revenue. The only revenue that matters. And why pay TSMC a huge chunk of your margins when you dont have to.

And probably 90+% of all global MPU profits as well.

About the only point that it would make sense for Intel to go fabless is when their wafer volume drops below the point at which they can fill a fab. Until that time happens there is no point in going fabless. To put it in perspective, Intel currently operates 10 active fabs producing between 14nm to 65nm products all on 300mm wafers. TSMC operates a total of 4 300mm fabs...

 

[imported_ats]

I can see the issue, where is the volume going to come from to sustain the R&D beast? Would hate for Intel to go fabless, it worked out so well for AMD as we all know.
Couldnt Intel fab ARM cores on the previous gen node or node before that? Chew out 22 and 32nm ARM socs left and right?

Intel's leading edge silicon wafer volume dwarfs pretty much everyone else combined. If Intel can't afford the R&D, no fab is going to be able to either.

Intel processes are too expensive for that. There's a reason why SoFIA was made at TSMC.

Intel's process more expensive? Not likely. Intel's process tend to be as cheap or cheaper than anyone else and have higher yield at a given node. Likely reason for TSMC for SoFIA was a combination of existing arrangements/designs and hard macro availability.

 

[Sheep221]

It's BS, and I only read the first 3 sentences.

 

[CHADBOGA]

After Intel's abominable efforts at Mobile, I don't think something like this can necessarily be automatically discarded.

Keeping in mind we would be looking at 5+ years before it happened, if it happens.

Intel's foundry attempts to date have been about as successful as their efforts in mobile.

 

[scannall]

If anything, they'd enter the foundry business. I don't see that as really likely but they could at least make some money in mobile that way.

 

[myocardia]

...

 

[norseamd]

So how are the IOT and embedded projects turning out for Intel right now?

 

[biostud]

If anything, they could always bring outsiders into their fabs if it would be profitable.

 

[ShintaiDK]

So how are the IOT and embedded projects turning out for Intel right now?

Intels IoT alone is larger than Samsungs entire logic foundry operations.

http://www.icinsights.com/news/bulletins/Apple-Drove-Entire-Foundry-Sales-Increase-At-TSMC-In-2015/

 

[MiddleOfTheRoad]

I can see the issue, where is the volume going to come from to sustain the R&D beast? Would hate for Intel to go fabless, it worked out so well for AMD as we all know.
Couldnt Intel fab ARM cores on the previous gen node or node before that? Chew out 22 and 32nm ARM socs left and right?

Nvidia is fabless. 3DFX had a fab.

Do I need to say anything more?

BTW, AMD is probably a much healthier company without a fab. If it weren't for the legacy wafer agreement, they'd already be in a strong market position.

 

[Sheep221]

After Intel's abominable efforts at Mobile, I don't think something like this can necessarily be automatically discarded.

Keeping in mind we would be looking at 5+ years before it happened, if it happens.

Intel's foundry attempts to date have been about as successful as their efforts in mobile.

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

 

[cytg111]

Nvidia is fabless. 3DFX had a fab.

Do I need to say anything more?

BTW, AMD is probably a much healthier company without a fab. If it weren't for the legacy wafer agreement, they'd already be in a strong market position.

I dont see Bulldozer+ being competitive given another fab, even Intels, do you? I guess that they're into WSA mess cause they was in dire need of cash and that was the only way to get it, - so from that we can deduce that if AMD had more money it would be a healthier company. Sure, ill buy that logic .

 

[poofyhairguy]

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

I thought this concept was proven to be a fairy tale?

 

[norseamd]

I thought this concept was proven to be a fairy tale?

What I am not getting is how using the word combined makes any difference other than to make Intel seem more impressive. I was not aware that the "technology" in the fabs was summable.

 

[Sheep221]

I thought this concept was proven to be a fairy tale?

It didn't
Intel does make good x86 CPUs, which means the most advanced integrated circuits and at rate no other company would do.

 

[Roland00Address]

I thought this concept was proven to be a fairy tale?

Intel easily has more advanced process (as in on the most leading node) wafers than the other fabs combine, that said total volume of wafers and sales of the other fabs catch up for the other fabs do not mind selling tech that is not on the most recent node.

Intel's advantage in technology and properties of the node they build their chips at though has been shrinking and it is far less than it used to be.

 

[witeken]

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.

 

[Qwertilot]

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

Sorry to be pedantic, but isn't that actually the same thing?

If the foundry part went full on foundry, and span off the CPU business, then you're left with a fabless CPU company