GlobalFoundries may subcontract Apple A-series chips for Samsung at New York facility

Posted:
in General Discussion edited January 2014
Rather than serving as an alternative to Samsung for production of Apples A-series chips, GlobalFoundries may just serve as a subcontractor for Samsung itself in producing chips for the iPhone and iPad, according to a new report.

A7
Apple's latest A7 SoC. | Source: Chipworks


Arik Hesseldahl of AllThingsD poured some cold water on the latest Apple silicon rumors on Tuesday when he revealed that Samsung is working on a deal that would have GlobalFoundries pick up additional work "on an as-needed basis." If the deal works out, Samsung would remain the sole provider of chips to Apple, and GlobalFoundries would simply handle "flex capacity" as a Samsung subcontractor.

"This is a long-standing industry practice under which a chip manufacturer pays to occasionally use another company's factories when demand on their own factory is running higher than they would like, and they need a little help," Hesseldahl explained.

The report refutes earlier speculation that GlobalFoundries would become a new chipmaking partner for Apple directly. That led some to believe that the company's New York-based operations would compete with Samsung and enter into Apple's supply chain.

To date, Samsung has been the sole supplier of A-series chips for Apple's iPhone and iPad, despite the fact that the South Korean company is also Apple's biggest rival in the consumer electronics space. Samsung produces custom chips such as the A7 found in the iPhone 5s, iPad Air and Retina iPad mini at its foundry in Austin, Tex.

GlobalFoundries has a chip production site in Malta, New York, which has been rumored to be in the running to produce chips for Apple for some time. But if a deal is finalized with GlobalFoundries, it would appear that it would be through Samsung, though Apple would give final approval on such a decision.

Apple has been long rumored to be working on a chipmaking deal with Taiwan Semiconductor Manufacturing Co., with the latest reports claiming the company could share production of Apple's anticipated "A8" chip in 2014. However, such rumors of an imminent deal with TSMC have persisted for years, and have yet to become a reality.

Comments

  • Reply 1 of 15
    Makes me wonder if Sammy is trying to make room for their own requirement of 64-bit chips.
  • Reply 2 of 15
    Quote:

    Originally Posted by thegreatbosan View Post



    Makes me wonder if Sammy is trying to make room for their own requirement of 64-bit chips.

    Thinking might also give  GlobalFoundries a chance to "practice" with Apple chips

  • Reply 3 of 15
    paxmanpaxman Posts: 4,729member
    Would it not make sense for Apple to run its own foundry?
  • Reply 4 of 15
    shompashompa Posts: 343member

    1) Samsung does not need more "space" of its own 64bit chips. They have enough wafer quantity to produce everything they need. 

     

    2) You can't take a chip design from Samsung and just start produce them at a GloFo foundry. You need a new tapeout, testwafers, verifying and probably re-spins. 

     

    According to Semiaccurate.com Apple already have a foundry/plan. Its behind a paywall, so I dont have any details. Semi are usually correct about foundry rumours. 

     

    Same with all that say that Intel should produce Apple A series Soc. There its a redesign Apple have to do on its socs. And people forget how big customer Apple is. Intel just have 2 22nm fabs and Apple would need 1 fab 24/7 just for their own chips.

     

    But its strange that iPhone and iPad uses the same SoC. iPad usually get a better SoC with more graphic cores. For Apple its a huge saving just using one SoC design. 

    If Apple had 2 different designs, they could design one for GloFo and one for Samsung.

     

    I personally have for years said that Apple should stop wasting money on dividend and share buyback. They should instead buy/build a high end foundry. Samsungs whole success story is because they have had Apples road map/chip designs since iPhone1. Samsung knows what kind of SoC Apple is going to use at least a year before the general public does. 

     

    The interesting thing about Apple is that they dont spend too much money on research. Apple however spends enormous money on tooling. For example paying sharp 500 millions to ramp up IGZO displays that made the iPad Air possible. Apple even outspend Intel on tooling and Intel have by far the best foundries in the world 1-1.5gen ahead of the competition. 

     

    And what happened to Apples own Flash/SSD? They bought Anobit years ago. 

     

    The reason why Samsung and Apple are the only two successful tablet/phone companies are: Samsung have their own fabs and can get all parts cheaper. Apple prepays billion to get parts cheaper. This is evidence that the idiotic outsourcing needs to die and Apple with all its money should manufacture all its own stuff. The A series SoC is a huge competitive advantage over all other ARM based/user companies.  

     

    1) Buy a foundry/built one

    2) Buy sharp LCD plant

    3) Sell developer versions of iOS stuff. Unsubsidized units without support and instead Root access to the user

    4) Sell Android iPhones. The money is in the hardware. If Apple can sell 10% more phones its billions in profit

    5) Dump Azure and make iCloud work. You can never outsource key business. 

    6) Gaming initiative for iOS and allow app installation on AppleTV.

    7) Cheap USB powered AppleTV for Airplay and that kind of stuff.

    8) Full A7 gaming version of AppleTV for gaming.

    9) Make Ive CEO. Tim Cook is not good. He accept to made bad products/mistakes. 

  • Reply 5 of 15
    Quote:

    Originally Posted by thegreatbosan View Post



    Makes me wonder if Sammy is trying to make room for their own requirement of 64-bit chips.

    If Apple start selling more devices then expected, their manufacturers will need to up production volume to keep up. Subcontracting things out like this means that you don't have to worry about leaving huge overheads in case Apple suddenly need several million more chips.

  • Reply 6 of 15

    I've never read so many dumb speculations. First of all, GF isn't subcontracting for Samsung. The second fastest growing foundry in the world isn't a subcontractor.

     

    Secondly, Apple won't buy a damn foundry, especially one that is private and deeply in the black who has extended plans of nearly $20 Billion in build out.

  • Reply 7 of 15
    gatorguygatorguy Posts: 24,211member
    I've never read so many dumb speculations. First of all, GF isn't subcontracting for Samsung.
    After doing their own investigation AllThingsD thinks that's just what the story really is: Global Foundries will be contracted by Samsung for the A7 overflow. Doesn't mean they're 100% correct but it makes more sense than Samsung helping GF to replace them as an Apple supplier.

    Perhaps you aren't aware that Globalfoundries and Samsung, along with IBM, already have a formal agreement to collaborate on certain projects and have for more than a decade. GF and Samsung aren't the fierce competitors you may imagine they would be. It may sound dumb at first glance but I think AllThingsD is probably accurate.
    http://www.commonplatform.com/

    EDIT: There's an article here explaining in better detail the intricate relationship between GF, Samsung and IBM. After reading it I wish there were more techs that saw the value in cooperation to better move the industry forward.
    http://techreport.com/review/24343/ibm-globalfoundries-and-samsung-offer-a-glimpse-of-chipmaking-future
  • Reply 8 of 15
    Quote:

    Originally Posted by shompa View Post

     

    2) You can't take a chip design from Samsung and just start produce them at a GloFo foundry. You need a new tapeout, testwafers, verifying and probably re-spins. 

     


     

    Samsung, IBM and GlobalFoundries are part of Common Platform. Their process is supposed to be identical. There is some requalification work involved, but it's far less than if they went to TSMC.

  • Reply 9 of 15
    wizard69wizard69 Posts: 13,377member
    shompa wrote: »
    1) Samsung does not need more "space" of its own 64bit chips. They have enough wafer quantity to produce everything they need. 
    Actually the evidence points the other way. Some years ago when demand from Apple was much lower, apple was taking 80% of Samsungs capacity. They then expended the plant roughly doubling capacity but that was years ago now. today Apples demands for chips is massive, i would not be surprised at all to find out that they are bumping up avians maximum capacity again.

    2) You can't take a chip design from Samsung and just start produce them at a GloFo foundry. You need a new tapeout, testwafers, verifying and probably re-spins. 
    You can't do that directly but tho isn't the big hurdle it might be going to another plant. Global is part of the same consortium that Samsung is part of. As such the tools are very similar so we aren't talking scaling Everest here.
    According to Semiaccurate.com Apple already have a foundry/plan. Its behind a paywall, so I dont have any details. Semi are usually correct about foundry rumors. 
    The sites name says it all.
    Same with all that say that Intel should produce Apple A series Soc. There its a redesign Apple have to do on its socs. And people forget how big customer Apple is. Intel just have 2 22nm fabs and Apple would need 1 fab 24/7 just for their own chips.
    Yep, Intel is a lone wolf here. However if deans for Intels processors continues to shrink they ill have lots of excess capacity.
    But its strange that iPhone and iPad uses the same SoC. iPad usually get a better SoC with more graphic cores. For Apple its a huge saving just using one SoC design. 
    I actually thought this was strange also but the reality is they roughly doubled performance. That really isn't that bad and as you note it saves them money. Howeveer it does create a issue with IPad only having roughly the performance of a cell phone.
    If Apple had 2 different designs, they could design one for GloFo and one for Samsung.
    Which would be stupid. You would want to keep the designs as similar as possible across manufactures so that upsets at one plant can be covered by the other.
    I personally have for years said that Apple should stop wasting money on dividend and share buyback. They should instead buy/build a high end foundry. Samsungs whole success story is because they have had Apples road map/chip designs since iPhone1. Samsung knows what kind of SoC Apple is going to use at least a year before the general public does. 
    If Samsung really had the ability to leverage any such knowledge they would be doing far better at SoC design and wouldn't be using the likes of Qualcomm and other processors. As for Apple building a foundry I'm not convinced that it is a smart move.

    The interesting thing about Apple is that they dont spend too much money on research.
    R&D is over a billion a quarter.
    Apple however spends enormous money on tooling. For example paying sharp 500 millions to ramp up IGZO displays that made the iPad Air possible. Apple even outspend Intel on tooling and Intel have by far the best foundries in the world 1-1.5gen ahead of the competition. 

    And what happened to Apples own Flash/SSD? They bought Anobit years ago. 
    Haven't you seen the new SSD's in the Mac Pro and other machines. I'm willing to bet there is a lot of Apple/Anobit IP in those controller chips. After all that is what Anobit offered to Apple IP related to the control of flash memories. People seem to think that Anobit was a flash memory maker which has never been true.
    The reason why Samsung and Apple are the only two successful tablet/phone companies are: Samsung have their own fabs and can get all parts cheaper. Apple prepays billion to get parts cheaper. This is evidence that the idiotic outsourcing needs to die and Apple with all its money should manufacture all its own stuff. The A series SoC is a huge competitive advantage over all other ARM based/user companies.  
    So what does Apple do with factories that are no longer suitable for its current hardware? Sometimes the economics just don't make sense. In the cue of a foundry or LCD plant I think it is hard to justify at all. for Apple to keep up they would have to constantly be disposing of plants at significant losses. Apple gets barely a year out of a process node.

    1) Buy a foundry/built one
    2) Buy sharp LCD plant
    3) Sell developer versions of iOS stuff. Unsubsidized units without support and instead Root access to the user
    4) Sell Android iPhones. The money is in the hardware. If Apple can sell 10% more phones its billions in profit
    5) Dump Azure and make iCloud work. You can never outsource key business. 
    6) Gaming initiative for iOS and allow app installation on AppleTV.
    7) Cheap USB powered AppleTV for Airplay and that kind of stuff.
    8) Full A7 gaming version of AppleTV for gaming.
    9) Make Ive CEO. Tim Cook is not good. He accept to made bad products/mistakes. 
  • Reply 10 of 15
    wizard69wizard69 Posts: 13,377member
    konqerror wrote: »
    Samsung, IBM and GlobalFoundries are part of Common Platform. Their process is supposed to be identical. There is some requalification work involved, but it's far less than if they went to TSMC.

    Well not identical but extremely similar. Samsung has optimized for low power while IBM and GF went for performance. On top of all of that you have Apples IP, via Intrisity, in Samsungs production process. So there might be more than zero effort required here, but it is still far less than going to TSMC.
  • Reply 11 of 15
    slurpyslurpy Posts: 5,384member
    Quote:
    Originally Posted by shompa View Post

     

    1) Buy a foundry/built one

    2) Buy sharp LCD plant

    3) Sell developer versions of iOS stuff. Unsubsidized units without support and instead Root access to the user

    4) Sell Android iPhones. The money is in the hardware. If Apple can sell 10% more phones its billions in profit

    5) Dump Azure and make iCloud work. You can never outsource key business. 

    6) Gaming initiative for iOS and allow app installation on AppleTV.

    7) Cheap USB powered AppleTV for Airplay and that kind of stuff.

    8) Full A7 gaming version of AppleTV for gaming.

    9) Make Ive CEO. Tim Cook is not good. He accept to made bad products/mistakes. 


     

    I just lost a good amount of brain cells reading that. 

     

    Re 9 - there's significant debate whether the iOS7 design, which Ive was in charge of, was even an improvement. The guy is a hardware design genius. But what the hell makes you think he knows anything about being  a CEO, and what it takes to run an entire company? There's a reason SJ didn't ask him to be CEO. Also, why the **** would you want him to take on all the responsibilities of a CEO, which he has no interest in- don't you think he's best suited to be in the position he's in- which is to focus on design? This "make Ive CEO" meme is the most lazy, intellectually empty, and thoughtless, meme I've ever heard. 

  • Reply 12 of 15
    Quote:

    Originally Posted by wizard69 View Post



     it does create a issue with IPad only having roughly the performance of a cell phone.

     

    You are seeing it backwards. The iPhone has the power of a tablet, it's just under-clocked, as far as CPU and GPU go, to save battery.

  • Reply 13 of 15
    Quote:

    Originally Posted by wizard69 View Post





    Well not identical but extremely similar. Samsung has optimized for low power while IBM and GF went for performance. On top of all of that you have Apples IP, via Intrisity, in Samsungs production process. So there might be more than zero effort required here, but it is still far less than going to TSMC.

    http://www.globalfoundries.com/newsroom/2011/20110830_Samsung.aspx

     

    Quote:


     The companies are proving the collaborative value of a synchronized platform by working with several customers to optimize processes and tooling for both the low-power and high-performance 28nm HKMG technologies. The synchronization process helps ensure consistent production worldwide, enabling customer chip designs to be produced at multiple sources with no redesign required, leveraging the customers’ design investments.


    From their press release, Samsung and GlobalFoundries both have a low power process and a high performance/low leakage process and both processes part of Fab Sync so production can move between Samsung and GlobalFoundries with no redesign. Given the ease of bringing GlobalFoundries on stream I wonder what the advantages are of pushing for TSMC as alternate supplier to Samsung?

  • Reply 14 of 15
    ksecksec Posts: 1,569member

    Everything Wizard said.

     

    Semiaccurate.com Are usually dead wrong about everything.

     

    Basically Samsung cant handle the A7 load, which for the first time ever is being simultaneously used by three Apple Products, the iPad Air, iPad Mini Retina, and 5s.

  • Reply 15 of 15

    Based on GF's performance in the last 3 years in delivering products and nodes on time and at the yields needed, it's good Apple is using them only for overflow. The contractual arrangement AMD has with GF from their spin-off is probably one of the biggest, if not the biggest, nooses around their neck.

Sign In or Register to comment.