Who will Apple partner with at 20nm: TSMC, Intel, Both, or Neither?

[Daniel Nenni]

Ever since Intel announced plans to offer foundry services at 22nm speculation has run rampant on whether Apple will move from Samsung to Intel or TSMC. I think it is a given Apple will, at a minimum, second source wafers as a result of the product conflict with Samsung. Second sourcing also gives Apple bargaining power and the opportunity to work with leading edge silicon.

The question is who will win Apples 20nm business? I must recuse myself but let us take a vote and see if the crowd is right once again.

D.A.N.

<script src="//platform.linkedin.com/in.js" type="text/javascript"></script>
<script type="IN/Share"></script>

 

[otistd]

D.A.N.
One thing about a second source is you would like to believe if the primary source goes down you can get enough extra capacity to (at least partially) offset the loss. Does Intel have enough extra capacity to take on a major Foundry customer like Apple? Granted Apple is already a big Intel customer, but how would the A6 (?) compete with Intel's own processors for capacity- and what about cost structure? I guess what I'm asking really is "how serious do you think Intel is about the foundry business- are they just testing the waters or ready to take the plunge?" and how does that fit into the "fabless model is dead"?

 

[Daniel Nenni]

Second sourcing semiconductor designs at the smaller geometries is not as easy as it sounds. There is no compatibility between TSMC 20nm, Intel 20nm, and Samsung 20nm, so you cannot just "move" them over. What I expect Apple to do is to have different products at the different foundries to get better pricing and capacity. A6 at Samsung, A7 at TSMC etc... Samsung has built two fabs just for Apple and is working on a third. TSMC will do the same.

Intel has said they will double 22nm capacity by the end of 2013. I do not expect the PC market to double so that capacity is for Atom SoCs. The Intel foundry business is in support of the Atom and mobile market so yes this is serious. Over the next 5 years Intel will work with strategic foundry customers on Atom based designs. Once successful Intel could certainly compete with the pure-play foundries but I highly doubt it. They have much higher margin business to pursue.

D.A.N.

 

[otistd]

OK, so TSMC clearly wins if it's between Intel and TSMC- unless Apple decides to do an Atom-based design. But what about GloFo? They could offer the best compatibility with Samsung as a true second source. Or, at least that is the claim from the CP- do you disagree? Or just think that TSMC will win on price and technology? It looks like you don't think GloFo is a serious option from the way you phrased the poll.

 

[Daniel Nenni]

OK, so TSMC clearly wins if it's between Intel and TSMC- unless Apple decides to do an Atom-based design. But what about GloFo? They could offer the best compatibility with Samsung as a true second source. Or, at least that is the claim from the CP- do you disagree? Or just think that TSMC will win on price and technology? It looks like you don't think GloFo is a serious option from the way you phrased the poll.

I think GloFlo is (1) year behind TSMC at 28nm. I believe they will trail TSMC at 20nm as well. Intel has done us all a favor by challenging TSMC at the process level. TSMC will race to catch them, believe it. The issue with price is how low will Intel go if/when supply outpaces demand?

D.A.N.

 

[LinkedIn]

INTEL will not build ARM chips. (Full Stop) (.) (lol) But Intel will HELP Apple with "the right commercial agreement" they will not build chips and be sucked by Apple,
Let Apple go to TSMC, intel do not need it if they cannot make the 60% GM. Apple hitting a wall,Tim Cook might talent of surviving and bend to Intel.

 

[LinkedIn]

With TSMC, there are lots of low cost third-party IP that Apple can fetch for its Ax design for pure low cost and time to market rush. Once Intel starts to release its PDK to outside design houses and more IP available, then Intel would be more attractive due to its technology power over TSMC.
Posted by Joshua Lu

 

[Daniel Nenni]

With TSMC, there are lots of low cost third-party IP that Apple can fetch for its Ax design for pure low cost and time to market rush. Once Intel starts to release its PDK to outside design houses and more IP available, then Intel would be more attractive due to its technology power over TSMC.
Posted by Joshua Lu

I agree, IP is key for the fabless guys. The ARM ecosystem as an example. Intel has made it clear they will NOT be an open foundry and will only work with strategic partners on Atom designs. This makes complete sense to me. Competing head-to-head with TSMC does not make sense at all. Just my opinion of course. If Apple uses Atom that would be a strategic foundry partner for Intel, absolutely.

 

[LinkedIn]

Daniel, this is my personal opinion.

Apple and Intel are two Titans fighting on very same market, fighting to delight you, me and the world with their products and features.
Intel is giant IT company, with 100% focus on the IT market segment, Desktops, Notebooks, Netbooks, Ultrabooks, Thin Clients, Servers, Embedded Systems, Data communication, Networking, Digital convergence, Softwares....

And with such profile and leadership position on these market segments makes Intel more likely to be a rival than partner.
Anyway, we never know what's is going on at the back stages, do we?

My best,
Posted by Osnir Gracia

 

[Daniel Nenni]

Daniel, this is my personal opinion.

Apple and Intel are two Titans fighting on very same market, fighting to delight you, me and the world with their products and features.
Intel is giant IT company, with 100% focus on the IT market segment, Desktops, Notebooks, Netbooks, Ultrabooks, Thin Clients, Servers, Embedded Systems, Data communication, Networking, Digital convergence, Softwares....

And with such profile and leadership position on these market segments makes Intel more likely to be a rival than partner.
Anyway, we never know what's is going on at the back stages, do we?

My best,
Posted by Osnir Gracia

I agree. Apple will be wary of Intel as a competitor from their experience with Samsung. Intel will also want Apple to switch from ARM to Atom which may be a problem.

I don't think Intel will really compete with TSMC as a foundry but it is certainly a healthy competition for the rest of us in the semiconductor ecosystem and fun to watch!

D.A.N.

 

[tupper9999]

Daniel,

Do you have any sense for how difficult the switch from ARM to Atom may be? Is this the PowerPC to X86 move all over again and if so, what sort of issues/bottlenecks/timeframes would you put on this? Also, what would have to happen at the application level to make sure that Apple doesn't screw up the ecosystem they've built up until this point? thanks, Chuck

 

[LinkedIn]

Daniel, this is my personal opinion.

Apple and Intel are two Titans fighting on very same market, fighting to delight you, me and the world with their products and features.
Intel is giant IT company, with 100% focus on the IT market segment, Desktops, Notebooks, Netbooks, Ultrabooks, Thin Clients, Servers, Embedded Systems, Data communication, Networking, Digital convergence, Softwares....

And with such profile and leadership position on these market segments makes Intel more likely to be a rival than partner.
Anyway, we never know what's is going on at the back stages, do we?

My best,
Posted by Osnir Gracia

I agree that TSMC has good chances but here is another theory:
What would prevent Apple from moving to x86 in the tablet market and use Intel foundry?

- Apple have the SW infrastructure for supporting x86 already and they already proved their capabilities to switch between architectures (PowerPC, ARM versions and x86).
- Moreover, Android will have many SW compatibility issues in moving to x86 - can be great differentiation for Apple to use x86.
- Intel is looking for ways to step into the tablet market - who's better partner than Apple for that purpose?

I'm not saying that it's going to happen but certainly 'never say never' rule applies here.

Mordi Blaunstein

 

[LinkedIn]

I agree that TSMC has good chances but here is another theory:
What would prevent Apple from moving to x86 in the tablet market and use Intel foundry?

- Apple have the SW infrastructure for supporting x86 already and they already proved their capabilities to switch between architectures (PowerPC, ARM versions and x86).
- Moreover, Android will have many SW compatibility issues in moving to x86 - can be great differentiation for Apple to use x86.
- Intel is looking for ways to step into the tablet market - who's better partner than Apple for that purpose?

I'm not saying that it's going to happen but certainly 'never say never' rule applies here.

Mordi Blaunstein

Mordi,
This is true and it can happen.
As I have mentioned, no one really knows what happens in the backstage, behind the scenes of big business, or negotiations between them.

Intel does not have the tablet, and if Apple would adopt the x86 architecture, theoretically, would be good for Apple, but much better for Intel, because they would be participating indirectly (or directly?) of the huge global market of tablets.

One thing we know for sure, both Intel and Apple have the same giant competitor, Samsung, and it makes sense they join in an effort to defeat the common rival.
Anyway, this whole very interesting discussion and opinions, are conjectures, and only time will bring the truth, then we can only, wait and see.

Apple, Intel, Samsung, Microsoft, Google among few others, are like very big, hungry and dominant dogs fighting for the leadership and space, in a very small room - the WORLD, disputing the same food - the MONEY (yours, mine and all people).


Osnir Gracia

 

[Xtoa]

This vote should be an easy win for TSMC. What am I missing?

 

[simguru]

...
- Moreover, Android will have many SW compatibility issues in moving to x86 - can be great differentiation for Apple to use x86.
...

Mordi Blaunstein

Given that there is an Android emulator that runs happily on my X86 laptop, and Android is just a GUI running on top of Linux, it seems unlikely that there are many issues in moving between X86 and ARM for Android.

Atom is still a little on the power-hungry side for the mobile market, but maybe that'll be fixed in a process node or two.

I expect Android will win out anyway regardless of platform since it is cheaper, and Apple are restrictive about what you can do on their platform - i.e. you can make Android/Linux do lots of stuff straight out of the box that Apple want you to pay extra for.

 

[jsn_nardllc]

Hey Dan, Intel and Altera had this product where an Intel Atom die and Altera die was integrated into one package. Any thoughts on further collaborations in this area? If Altera gets access to Intel's latest process technology for their product line, that would be a real leg up for them.

Intel has cut deals with a couple of FPGA startups. Did it involve Intel's latest processes?

Best Regards,

Jim

 

[Daniel Nenni]

Hey Dan, Intel and Altera had this product where an Intel Atom die and Altera die was integrated into one package. Any thoughts on further collaborations in this area? If Altera gets access to Intel's latest process technology for their product line, that would be a real leg up for them.

Intel has cut deals with a couple of FPGA startups. Did it involve Intel's latest processes?

Best Regards,

Jim

Yes Intel will be making FPGAs at 22nm and below. FPGAs are used to ramp new processes so this makes complete sense. Altera and TSMC are very close and Altera is generally first to come off a new TSMC process, but now SoCs are driving wafer volume so new processes will be optimized for mobile. For example, Altera uses the HP version of 40nm and 28nm. TSMC will only have one 20nm version and that is 20nm SoC (mobile). Altera may look elsewhere but I doubt it. The transition will be expensive and in my opinion Intel as a foundry is still an unknown.

I can't remember a more interesting time in the semiconductor ecosystem, Intel sure has stirred things up a bit!

D.A.N.

 

[omahesh]

Intel is really struggling at 22nm production itself, where as TSMC has succeeded in the production of 28nm. TSMC guys soon they will come up with an advancement than intel people.

TSMC can acquire apple's 20nm contract.

Mahe

 

[Boris]

A good question - who knows?

A possible switch to Intel is perhaps more likely in 14nm node - for Intel in 2014.
So that might happen for Apple at Intel in 2015/16. By 2016 Intel will also be in 450mm wafers.

Very advanced nodes apparently bring benefits in power and performance - 450mm might introduce cost benefits again.
And Intel, just like TSMC and Samsung, will need a huge volume to increase yields on 450mm.

And - there is the issue of packaging. Who will be best able to package 3D TSV processors and memories?
And - will FD-SOI be required for 14nm Fin-FET? Who will make 450mm FD-SOI - Soitec?