Rumor: Apple buys into chip fab, plans to build its own silicon

2

Comments

  • Reply 21 of 44
    richard getzrichard getz Posts: 1,142member

    Quote:

    Originally Posted by AaronJ View Post


    I don't buy this for a second.  I don't see how this would be to be Apple's advantage, at all.  It seems like getting involved in something that doesn't really suit them.


     


    What am I missing?



     


    The biggest advantage is technology secrets staying secret until release. Currently Samsung has new design concepts probably a year before release. They can't copy the design, but they can learn from it and incorporate similar features. No one knows what they don't know, so when you have someone else showing you design concepts that may be years ahead of where you were thinking, that will give you enlighten you to knew concepts. 


     


    Example: original iPhone and it taking 2-3 years for competition to start catching up. If Apple could do that with processors, that will put them 2-3 years ahead of Samsung and others. 

  • Reply 22 of 44
    sockrolidsockrolid Posts: 2,789member


    Makes sense.  Apple would be able to actually own the means of production of the heart of all of their mobile products (and TV products.)  And it's easier to prevent future product leaks from people whose salary you directly control.  Adds a new meaning to "doubling down on secrecy," doesn't it?

  • Reply 23 of 44
    sockrolidsockrolid Posts: 2,789member


    Originally Posted by Richard Getz View Post


     


    The biggest advantage is technology secrets staying secret until release. Currently Samsung has new design concepts probably a year before release. They can't copy the design, but they can learn from it and incorporate similar features. No one knows what they don't know, so when you have someone else showing you design concepts that may be years ahead of where you were thinking, that will give you enlighten you to knew concepts. 


     


    Example: original iPhone and it taking 2-3 years for competition to start catching up. If Apple could do that with processors, that will put them 2-3 years ahead of Samsung and others. 



     


    Especially if Apple develops (and patents) original processor fabrication technology and/or chip design.


    I feel the need to quote a classic sci-fi action flick here...


     


    "But it gave us ideas, It took us in new directions... things we would never have thought of. All my work is based on it."


    - Miles Dyson (Joe Morton, "Terminator 2: Judgement Day," 1991)

  • Reply 24 of 44
    jessijessi Posts: 302member
    Building a fab is just a question of money, but making that fab produce compelling chips requires process expertise. This is why Apple would want to buy TSMC, or if they aren't willing to sell, UMC. This is also why Intel would be smart to set up a joint venture with Apple, much like they did with micron in flash. But intel is not smart.

    So, Apple buying a large chunk, possibly controlling interest, in UMC makes a lot of sense.

    It will give them a great deal of freedom to innovate, and it will be harder to steal their innovations.
  • Reply 25 of 44
    mechanicmechanic Posts: 805member

    Quote:

    Originally Posted by AjbDtc826 View Post



    I thought they bought that chip maker out in Isreal a few years ago, sounds to me they've been looking into this for some time now.


     


    You mean Anobit, and yes they did buy them out, there also building a campus in israel and hiring a ton of microelectronic engineers.  Apple already uses anobit's memory controller in the iPhone and iPad and iPod, that is the main reason they bought them.  They have one of the most power miserly, and fastest memory controllers available. The key to anobits controller is that it extends flash memory's life.


     


    Here is a good article on what they got from the Anobit purchase:  http://news.cnet.com/8301-13924_3-57346212-64/anobit-acquisition-keeps-apple-ahead-in-flash-memory/#!


     


    A quote from the article 


     


    Quote:

    Anobit has developed a memory signal processor, or MSP, that is able to manage very high bit error rates and extend the life of flash memory devices

  • Reply 26 of 44
    ajbdtc826ajbdtc826 Posts: 190member
    Thanks to all who responded to post, good material!
  • Reply 27 of 44
    ksecksec Posts: 1,569member
    It makes zero sense.

    UMC is not, and has never been at node leading Fabs.

    I have been saying this for far too many times. You cant switch Chips from One Fabs to Another like Copy and Paste. It doesn't work. Even If the Fabs are all based on IBM Fabs Alliance ( Global Foundry and Samsung ) it will still takes a few months to get things work and yield.

    You cant try to have two different supplier of Fabs on leading node. Because they will very likely have a different timing schedule on it.

    Owning Your Own Fabs means keeping it up with capacity, node, yield, upgrades etc. As if Apple's Operation does not have enough to deal with already?

    As much as the analysis wants you to believe Fabs making big margin, these 60% margins from 28nm TSMC will not last. And only happened due to a perfect storm scenario. Price War will drive this back down once we reach 16 / 14 nm which everyone ( Intel, Samsung, TSMC, Global Foundry ) is well prepared for.

    And if you are still questioning the Fat margins that Apple should bring this in house. You should also consider the actual Margins over the lifetime of node. Only leading node are gaining the highest margin are the 60% were only driven by mobile demand to TSMC that far exceed their ( clients ) expectation. Two years later when new node arrive they will drop significantly. Fabs Business is an extremely competitive market.

    If Apple is really buying UMC I would likely think this has more to do in partner ship with TSMC where there has been rumors for years that TSMC may buy UMC. Sharing Tech and equipment from TSMC will be like working with TSMC but with a specific Fabs dedicated to Apple.

    But then those UMC Fabs close to TSMC are not cutting edge, still on 200mm wafers and will need another year or two to upgrade. And UMC will have far too Fabs that Apple doesn't want.

    So to conclude with my opening statement

    It makes zero sense.
  • Reply 28 of 44

    Quote:

    Originally Posted by cnocbui View Post


    Wow, TSMC will be happy.  They haven't even managed to turn away from the table where they signed the latest agreement before there was a dagger between their shoulderblades - if true.



     


    Explain please. Did Apple renege on their contract?

  • Reply 29 of 44
    qamfqamf Posts: 87member


    So, UMC isn't relevant.

    http://semiaccurate.com/forums/showpost.php?p=188908&postcount=39

     


     


    Quote:Hive


    Quote:Thrillseeker


    [this article on Apple Insider]

    Oh noes, they are putting words in the Semi-Accurate mouth! (and they ain't wrapped in bacon)



    It's time to unleash the granny!



    Even the paywall they got partially wrong - Student Level is enough for most to know how wrong UMC is.



     


    and note: the student one exists.  I am a subscriber to S|A (student level).

     


    I concur with Hive.



    -QAMF

  • Reply 30 of 44
    qamfqamf Posts: 87member

    Quote:

    Originally Posted by drblank View Post


    Heck, Apple has more than enough money to do it.  It typically costs a few billion for the plant/equipment.  Now all they have to do is buy Sharp and keep the IGZO panel technology to themselves.



    Fabs + research probably starting to hit close to 10 billion now.



    (that is 10 billion for construction + research + tools for a node).


    Quote:

    Originally Posted by whoda View Post



    First and foremost, I'd be suspicious of anything behind a $1000 a year paywall. They are obviously going to say anything that might get people to pay up. But I think we've all seen this coming for a while. They keep buying design shops, and they have hired an entire team from AMD, and they kept old Bobby around for a reason. I'm guessing something like this takes a lot of time to get up and running, so TSMC is probably just a short term solution, while building their own chips is eventually the goal. I think they should just buy up their supply chain, and bring it all in house!


    It is also behind a 100$/year firewall.... AI needs to fix the article .



    TSMC is the "Samsung is to damn against us to work with" option.  But yes, probably short term one.



    I would be surprised if Apple's fab gets up before 16/14nm node, which would be 2-3+ years (for mass production).


    Quote:

    Originally Posted by al_bundy View Post


    every company of apple's size has multiple suppliers


    even apple's current products use multiple suppliers for iphones, mac's, etc



    Huh?  not for a single SoC.



    Every single fab has different design specs and capacities, two A6 SoC's could have completely different power usages at the ~1.3Ghz stock in the iPhone 5.



    It is hard to source multiple fabs due to that.


    Quote:

    Originally Posted by cnocbui View Post


    Wow, TSMC will be happy.  They haven't even managed to turn away from the table where they signed the latest agreement before there was a dagger between their shoulderblades - if true.



    TSMC has all of Apples biz for at least 20nm (and I believe 20xm (ie the fusion of 20 and 16/14nm nodes))  That is what TSMC bid for, that is what they won.



    Cheers!

    ~QAMF

  • Reply 31 of 44
    xiao-zhixiao-zhi Posts: 112member
    Um, no. There would be no reason for Apple to buy UMC when they can get as much capacity as they need and better technology elsewhere. If they were going to buy into a fab it would probably be Global Foundries but I see little incentive for them to do so when they can just buy chips.
    [I]
    SemiAccurate[/I] seems [I]SemiDesperate[/I] for attention.
  • Reply 32 of 44
    xiao-zhixiao-zhi Posts: 112member

    Quote:

    Originally Posted by Richard Getz View Post



    I have been saying this for months and I am glad to hear they might be. This will give them an advantage not seen before. The ability to produce processors away from competitors view so they are first to market with advanced technology. If done correctly, this alone could put the iOS devices 2-3 years ahead of competition.



    Apple can start small here but doing most of the design work and small runs to work out issues and then move mass production to one of the bigger fabs. This allows that 1-2 year design process to be fully vetted before competition could even have the chance to see it.


    You have been saying for months Apple would buy UMC?  I bet.


     


    This is nonsense; UMC is not an attractive partner, would not be worth buying and this rumor will turn out to be more the the usual BS speculation.


     


    And dude, IC design does not "start small" and then get moved from fab to fab, but it's an amusing idea.

  • Reply 33 of 44
    qamfqamf Posts: 87member

    Quote:

    Originally Posted by Xiao-zhi View Post



    Um, no. There would be no reason for Apple to buy UMC when they can get as much capacity as they need and better technology elsewhere. If they were going to buy into a fab it would probably be Global Foundries but I see little incentive for them to do so when they can just buy chips.



    SemiAccurate seems SemiDesperate for attention.


    Have you read the article?  No.  I doubt it.  I don't know where anyone is getting UMC from.  I guess that it is a tag....  But so are Intel, and IBM.



    If you cannot read something, you have no grounds to criticize it on.



    And, it is not UMC.... because UMC would be stupid, as you pointed out.  The only thing that UMC has is FD-SOI on 28nm.... Apple is not looking to manufacture stuff on 28nm.





    Thanks for needless criticism,

    -QAMF

  • Reply 34 of 44
    richard getzrichard getz Posts: 1,142member

    Quote:

    Originally Posted by Xiao-zhi View Post


    You have been saying for months Apple would buy UMC?  I bet.


     


    This is nonsense; UMC is not an attractive partner, would not be worth buying and this rumor will turn out to be more the the usual BS speculation.


     


    And dude, IC design does not "start small" and then get moved from fab to fab, but it's an amusing idea.



     


    No, sorry you read my post as such. I have been saying for months that Apple needs their own fab. 


     


    Also, dude, the small was in effort compared to full production of their own chips. While design and processing may be more intensive, more critical, it is less labor than production. The point was before they jumped to tens of millions of processor production, they could work on design and process thus keeping the intellectual property hidden for as long as they could. 

  • Reply 35 of 44
    xiao-zhixiao-zhi Posts: 112member


    Since the story was about UMC and your comment lead with "I have been saying this for months and I am glad to hear they might be" my assumption is you refer to the story subject. 


     


    Personally, I don't think Apple needs it's own fab as an increasing amount of foundry capacity is becoming available including Global Foundries and Intel, both of whom have the ability to replace the role of Samsung should Apple really decide to drop them (but seems not). If you look at the history of Apple courting UMC and TSMC, this happened at a time when such capacity was short. UMC made a pass and their CEO is smoking what I don't know, TSMC didn't have the cap to make a big commitment but started qual trials with A6. Since then, the market has significantly changed with (a) the demise of PCs reducing the demand for Graphics chipsets and (b) the success of Intel promoting integrated graphics for notebooks.


     


    This makes UMC a real non-starter, technologically, and in a weak position, so ripe for rumors but not a very attractive target. Buying physical fabs is buying other people's problems (whatever they are) and UMC is long in the tooth.


     


    However, Apple talks with GF seem to have stalled. 


     


    You said:


     


    "Apple can start small here but doing most of the design work and small runs to work out issues and then move mass production to one of the bigger fabs. This allows that 1-2 year design process to be fully vetted before competition could even have the chance to see it."


     


    I guess we disagree. My experience is this model no longer works as designs must be engineered for specific processes and then debugged, which is a designer to fab proposition, and particularly the case with complex SoC designs. This was a viable model in the past for lower volume, less critical designs (I started in a captive fab) but really does not work for high volume leading edge designs. 


     


    Apple is not designing at an approximately 2 year cadence, they pretty much have to pick a fab site & process, and then use the time to design and debug and I don't see hand-off to new plants as a viable strategy.


     


    Of course, they could do that if they wanted: they have the cash to burn, at least for now.


     


    My vote for the dark horse is Intel. Established & trusted Apple supplier. Deep technology including leading R+D. Increasing capacity available. Interesting problem.


     


    Yes, we agree, 28nm is not interesting. Bogus claim.

  • Reply 36 of 44
    qamfqamf Posts: 87member

    Quote:

    Originally Posted by Xiao-zhi View Post


    Personally, I don't think Apple needs it's own fab as an increasing amount of foundry capacity is becoming available including Global Foundries and Intel, both of whom have the ability to replace the role of Samsung should Apple really decide to drop them (but seems not). If you look at the history of Apple courting UMC and TSMC, this happened at a time when such capacity was short. UMC made a pass and their CEO is smoking what I don't know, TSMC didn't have the cap to make a big commitment but started qual trials with A6. Since then, the market has significantly changed with (a) the demise of PCs reducing the demand for Graphics chipsets and (b) the success of Intel promoting integrated graphics for notebooks.


     


    This makes UMC a real non-starter, technologically, and in a weak position, so ripe for rumors but not a very attractive target. Buying physical fabs is buying other people's problems (whatever they are) and UMC is long in the tooth.


     


    However, Apple talks with GF seem to have stalled. 



    Yup, right on, except, what is the problem with buying a physical fab?  If this was a fab for 20nm (or above) I would agree, but process technology advancement beyond 14nm seems to be very slow.



    No reason why a 16/14nm fab wouldn't be a good buy.



    And when did talks with GF stall?


     


    Quote:

    Originally Posted by Xiao-zhi View Post


    Apple is not designing at an approximately 2 year cadence, they pretty much have to pick a fab site & process, and then use the time to design and debug and I don't see hand-off to new plants as a viable strategy.


     


    Of course, they could do that if they wanted: they have the cash to burn, at least for now.


     


    My vote for the dark horse is Intel. Established & trusted Apple supplier. Deep technology including leading R+D. Increasing capacity available. Interesting problem.




     


    No one is designing at a 2 year cadence, Intel isn't designing at a 2 year cadence anymore.



    Intel could do it, except that would Intel be accepting of ARM chips being made in their fabs.  And also the fact that Intel is a high-margin player, so is Apple.  Of the fabs I think Intel's "need" for high margins in the greatest.


     


    Quote:


    Originally Posted by Xiao-zhi View Post


    Yes, we agree, 28nm is not interesting. Bogus claim.



    Yes, just like this claim is Bogus:


    Quote:

    Originally Posted by Xiao-zhi View Post

    SemiAccurate seems SemiDesperate for attention.


     


    Complete bogus, all that S|A did was tag fabs.  Other sites choose a tagged fab, and flew with it.



    I say that having read the student level article.



    -QAMF



    EDIT: so far, S|A has been very accurate on things Apple is doing (or explained why they were wrong if wrong, like expecting 64-bit ARM to be ready much earlier then it will be)

  • Reply 37 of 44
    richard getzrichard getz Posts: 1,142member

    Quote:

    Originally Posted by Xiao-zhi View Post


    Since the story was about UMC and your comment lead with "I have been saying this for months and I am glad to hear they might be" my assumption is you refer to the story subject. 


     


    Personally, I don't think Apple needs it's own fab as an increasing amount of foundry capacity is becoming available including Global Foundries and Intel, both of whom have the ability to replace the role of Samsung should Apple really decide to drop them (but seems not). If you look at the history of Apple courting UMC and TSMC, this happened at a time when such capacity was short. UMC made a pass and their CEO is smoking what I don't know, TSMC didn't have the cap to make a big commitment but started qual trials with A6. Since then, the market has significantly changed with (a) the demise of PCs reducing the demand for Graphics chipsets and (b) the success of Intel promoting integrated graphics for notebooks.


     


    This makes UMC a real non-starter, technologically, and in a weak position, so ripe for rumors but not a very attractive target. Buying physical fabs is buying other people's problems (whatever they are) and UMC is long in the tooth.


     


    However, Apple talks with GF seem to have stalled. 


     


    You said:


     


    "Apple can start small here but doing most of the design work and small runs to work out issues and then move mass production to one of the bigger fabs. This allows that 1-2 year design process to be fully vetted before competition could even have the chance to see it."


     


    I guess we disagree. My experience is this model no longer works as designs must be engineered for specific processes and then debugged, which is a designer to fab proposition, and particularly the case with complex SoC designs. This was a viable model in the past for lower volume, less critical designs (I started in a captive fab) but really does not work for high volume leading edge designs. 


     


    Apple is not designing at an approximately 2 year cadence, they pretty much have to pick a fab site & process, and then use the time to design and debug and I don't see hand-off to new plants as a viable strategy.


     


    Of course, they could do that if they wanted: they have the cash to burn, at least for now.


     


    My vote for the dark horse is Intel. Established & trusted Apple supplier. Deep technology including leading R+D. Increasing capacity available. Interesting problem.


     


    Yes, we agree, 28nm is not interesting. Bogus claim.



     


    The problem with Intel, or anyone else, is that Apple's design will be in the hands of another manufacture just like Samsung. Even if they can't copy the design, they can learn from it. As you said 'dark horse is Intel' [taken out of context] and not the white horse. 


     


    I guess we disagree. My experience is this model no longer works as designs must be engineered for specific processes and then debugged, which is a designer to fab proposition, and particularly the case with complex SoC designs. This was a viable model in the past for lower volume, less critical designs (I started in a captive fab) but really does not work for high volume leading edge designs.


     


    I'll go with your experience, but hope that Apple still works a way to keep their intellectual property out of the hands of competition as long as they can. The sting of Samsung and Google is all to fresh to forget. 

  • Reply 38 of 44
    xiao-zhixiao-zhi Posts: 112member

    Quote:

    Originally Posted by QAMF View Post


    Yup, right on, except, what is the problem with buying a physical fab?  If this was a fab for 20nm (or above) I would agree, but process technology advancement beyond 14nm seems to be very slow.



    No reason why a 16/14nm fab wouldn't be a good buy.



    And when did talks with GF stall?


     


     


    No one is designing at a 2 year cadence, Intel isn't designing at a 2 year cadence anymore.



    Intel could do it, except that would Intel be accepting of ARM chips being made in their fabs.  And also the fact that Intel is a high-margin player, so is Apple.  Of the fabs I think Intel's "need" for high margins in the greatest.


     


    Yes, just like this claim is Bogus:


    Complete bogus, all that S|A did was tag fabs.  Other sites choose a tagged fab, and flew with it.



    I say that having read the student level article.



    -QAMF



    EDIT: so far, S|A has been very accurate on things Apple is doing (or explained why they were wrong if wrong, like expecting 64-bit ARM to be ready much earlier then it will be)



     


    A4 March 2010 -> (12 months) A5 March 2011 -> (18 months) A6 Sept 2012, so let's say recently they are turning 12-18 months. I do not count variants of a basic design.


     


    As for Apple buying UMC, considering the article is now 12 days old, I wonder why this has zero traction outside S|A and Apple Insider. Maybe I'll be proven wrong, but I seriously doubt Apple would burn cash on UMC just to buy other people's problems, they have more important things to do.


     


    Intel is a bit of a long shot, but they are now "soft" marketing foundry services and seem less that 100% confident in the handheld space, and is already on the record stating they could produce ARM designs. Fact is, Intel has huge, modern Fab capacity and are seeing serious erosion of desktop sales without corresponding gains of revenue in the tablet or phone space because Atom does not have the real estate or margin to replace lost PC demand. For at least 3 years they have fabbed custom silicon for Google (and now for Facebook) on the server side, so they are already halfway there. Clearly Intel does not plan on making it on desktops or notebooks alone much longer as they recently wound-up their desktop systems groups producing reference designs and transferred them to mobil product development. 


     


    Given Intel has very deep technology, too much capacity and a long relationship with Apple, there is a business case for both parties. And if we consider that Apple designers need to hitch their wagons to a fab process at the beginning of a design cycle, Intel brings a lot to the party that even Samsung and Global Foundries lack.


     


    My Take: UMC, no. TSMC, alternate generation, 2nd tier. Samsung, Yes. Intel, maybe. GF, maybe.


     


    If Apple buys UMC I'll buy lunch. 

  • Reply 39 of 44
    xiao-zhixiao-zhi Posts: 112member

    Quote:

    Originally Posted by Richard Getz View Post


     


    The problem with Intel, or anyone else, is that Apple's design will be in the hands of another manufacture just like Samsung. Even if they can't copy the design, they can learn from it. As you said 'dark horse is Intel' [taken out of context] and not the white horse. 


     


    I guess we disagree. My experience is this model no longer works as designs must be engineered for specific processes and then debugged, which is a designer to fab proposition, and particularly the case with complex SoC designs. This was a viable model in the past for lower volume, less critical designs (I started in a captive fab) but really does not work for high volume leading edge designs.


     


    I'll go with your experience, but hope that Apple still works a way to keep their intellectual property out of the hands of competition as long as they can. The sting of Samsung and Google is all to fresh to forget. 



     


    I understand your feelings about Samsung, but Apple and Samsung have a pretty deep and long relationship that might not be broken by recent events. In fact, I would expect them to continue doing lots of business but change the terms of engagement. I'm pretty sure Sumsung would like to continue fabing Ax chips in Austin, they have about $7Bn invested in the fab capacity used for these products and won't find other customers that easy.

  • Reply 40 of 44
    qamfqamf Posts: 87member

    Quote:

    Originally Posted by Xiao-zhi View Post


    As for Apple buying UMC, considering the article is now 12 days old, I wonder why this has zero traction outside S|A and Apple Insider. Maybe I'll be proven wrong, but I seriously doubt Apple would burn cash on UMC just to buy other people's problems, they have more important things to do.


     


    My Take: UMC, no. TSMC, alternate generation, 2nd tier. Samsung, Yes. Intel, maybe. GF, maybe.


     


    If Apple buys UMC I'll buy lunch. 



    Let me repeat this: Semiaccurate NEVER SAID ANYTHING about UMC.  They tagged it, because they were tagging foundries that had interesting technologies/furthest ahead in nodes.  UMC did get 28nm SOI when no one else did, aside from that, I dunno why else they could even be considered for that list.


     


    UMC isn't even mentioned in the article.


    Quote:


    Originally Posted by Xiao-zhi View Post


    Intel is a bit of a long shot, but they are now "soft" marketing foundry services and seem less that 100% confident in the handheld space, and is already on the record stating they could produce ARM designs. Fact is, Intel has huge, modern Fab capacity and are seeing serious erosion of desktop sales without corresponding gains of revenue in the tablet or phone space because Atom does not have the real estate or margin to replace lost PC demand. For at least 3 years they have fabbed custom silicon for Google (and now for Facebook) on the server side, so they are already halfway there. Clearly Intel does not plan on making it on desktops or notebooks alone much longer as they recently wound-up their desktop systems groups producing reference designs and transferred them to mobil product development. 


     


    Given Intel has very deep technology, too much capacity and a long relationship with Apple, there is a business case for both parties. And if we consider that Apple designers need to hitch their wagons to a fab process at the beginning of a design cycle, Intel brings a lot to the party that even Samsung and Global Foundries lack.




    Intel is currently on a quest to make sure x86 beats ARM, sadly, part of that quest means they need their process advantage.  Considering their low fab usage It could happen I guess.



    The issue is that Intel is a company that wants high margins, Apple is also.  Apple also is a master of the supply chain, able to get very cheap prices.



    The others, well, IBM would be interesting, but they wouldn't sell all their fabs (probalby would be what Intel would need lol)



    TSMC, GloFo, Samsung: One of those has an advantage that no one is looking at right now.



    -QAMF



    NOTE: SEMIACCURATE DID NOT SAY ANYTHING ABOUT UMC, SO STOP SAYING THEY DID

Sign In or Register to comment.