Analyst: Intel should open up fabs to Apple

The idea of Intel manufacturing Apple’s proprietary A-series chips isn’t new—the rumor mill broached the subject a year a and half ago. According to a Credit Suisse analyst quoted by Forbes, however, this hypothetical partnership could actually work.

Credit Suisse’s John Pitzer ran the numbers and found that, although Intel’s foundry business has a higher cost per wafer than TSMC, the world’s biggest contract foundry, Intel’s bottom line would benefit from a manufacturing partnership with Apple. Here’s the snippet from the Forbes piece:

His view: If Intel were to produce 100% of Apple’s demand at 14 nm the company generate an extra $5.8 billion in revenue, 47% gross margins (which is below the 60%-65% of the core business) but with operating margins of 35%, above the core of 28%-32%. He thinks that arrangement would produce an extra 28 cents a share in annual profits, with an ROI of 20%-25%.

Intel has opened up its fabs before. As Forbes points out, Intel Foundry Services builds chips for three little-known firms: Acrhnoix, Tabula, and Netronome.

Apple currently relies on Samsung to manufacture its chips, but Forbes says “the scuttlebutt is that Apple has been looking for alternatives.” Given the vicious litigation between the two companies, that’s hardly surprising.

A partnership with Intel might be beneficial to Apple, especially if Intel’s foundry business remains ahead of the rest of the industry. Pure-play foundries may be catching up, though. X-bit labs says GlobalFoundries is planning for early 14-nm tape-outs in late 2013. A story by The Register posted around the same time said Intel’s 14-nm process is “on track for production readiness around the end of next year.”

Comments closed
    • Game_boy
    • 7 years ago

    Intel do not enter <60% gross margin businesses even if they stand to profit from it, because of their institutional investors. Not happening.

      • NeelyCam
      • 7 years ago

      For the longest time Intel’s gross margins where closer to 50% – the 60+% margins are a new thing in the past five years or so.

      Investors were pretty happy with 50% margins back then – the stock was no worse then than it is now. In fact, it’s kind of crazy that Intel has boosted gross margins from 50% to 60% [i<]and[/i<] increased revenue massively in the past few years, yet the stock price hasn't moved; P/E has crashed instead.

    • stacey1x0pp
    • 7 years ago
    • jdaven
    • 7 years ago

    The race to 14 nm reminds me of the race to 1 GHz. It’s about time we have a good ole fashion technology contest.

      • NeelyCam
      • 7 years ago

      Race? If there is a race, it’s on who makes 14nm second

        • MadManOriginal
        • 7 years ago

        Second place? That’s just a fancy word for losing!

          • dextrous
          • 7 years ago

          If you’re not first, you’re last!

    • BobbinThreadbare
    • 7 years ago

    Yeah Intel should totally give up the high margin business of selling chips and let 3rd parties use their fabs to make whatever they want thus giving up a major advantage thatn Intel has.

    That totally makes sense.

      • The Dark One
      • 7 years ago

      Yeah, like others have said, why would they ever want to open up their fabs to a competing architecture? Intel’s big advantage is that their x86 chips are always a generation ahead in terms of process. There are already rumours of an ARM-derived desktop chip from Apple- why would Intel want to enable that?

    • tviceman
    • 7 years ago

    Industry analysts are short sighted and often do not have the entire picture.

    Apple is moving to eventually designing all of it’s own chips for both it’s phones, tablets, and computers. I’m guessing we are 3 years (or less) away from some or all of the Mac line running Apple designed ARM processors and Intel is pushed out of Apple products. Apple won’t agree to use Intel chips in their iPads unless they want to end up with an iPad with 40% more weight and 40% less battery life.

    If there was no-strings-attached with Apple getting access to Intel’sfabs, Intel would be opening themselves up to a competitor (and a very, very, very cash rich highly capable competitor). It would be tantamount to shooting yourself in the foot to avoid getting beat up in a fist fight. The idea of Intel making SoC’s for any of Apple’s products that already have Apple designed SoC’s is ludicrous. Apple won’t outsource what they are already doing as effectively as anyone else on the market.

    This rumor is about as dumb as the Intel / Nvidia merger that floated around the web last week.

      • homerdog
      • 7 years ago

      In one fell swoop Apple will design a viable competitor to Broadwell/Skylake, and find someone with manufacturing prowess comparable to Intel to produce it for them.

      I keed, but if they do choose to go down that road they should be prepared to be at a major competitive disadvantage for a long time, and likely never catch up to Intel.

    • anotherengineer
    • 7 years ago

    Apple has enough cash to make their own Fab if they wanted. Then they have complete control and don’t have to rely on anyone else.

    They could also make chips for other parties if they don’t use the full capacity.

      • NeelyCam
      • 7 years ago

      You can’t just “make” a fab. It takes massive amount of research and development. Apple has the cash, but not the expertise or time.

      If they want a fab in the USA, their best bet is to beg Samsung to give them one, buy GloFo or buy Intel. None of those options are likely

        • anotherengineer
        • 7 years ago

        The factory yes just build one, the node I agree with you, not exactly an off the shelf item. Doesn’t mean they couldn’t purchase a old 65nm node from GF or someone else, along with the brains.

        The cash can buy the expertise and the time, money talks.

        All I said is they have the cash, doesn’t mean they would want to.

          • willmore
          • 7 years ago

          Isn’t Fujitsu getting out of the fab business? I think they had an ice 65nm fab at one point. They already sold a 130nm fab to Microchip a few years back.

          • MadManOriginal
          • 7 years ago

          Buying a 65nm fab would only be useful for the building shell, the equipment would be too obsolete to be meaningful to Apple. It might still make sense since these buildings are pretty fancy but I still don’t see it happening. If Apple wanted a fab they would have done it already. I think they’re happy playing foundries off each other.

      • Horshu
      • 7 years ago

      They’d need to spend somewhere around $3-5 billion just to build the fab. They’d need to hire several engineers in different disciplines to figure out wafer production, license 3D transistor tech from someone (unless they invent some new way to get around leakage at the node they’d need to target). Then they would blow untold of amounts of cash and time trying to figure out how to get their yields up. If they’re lucky, by the time they figure everything out, the node the designed their fab around wouldn’t be obsolete. I just don’t see it being very cost-effective for them to go this route.
      Taking this into account as well as doubts as to whether Intel would even agree to manufacture ARM, they’d have several different better options:
      1. Buy TSMC or another foundry (VERY expensive, but at least they wouldn’t have to do it themselves)
      2. Start using Atom-based chips in their ARM-based offerings. Not as difficult as it sounds, given that they can already run their OS on x86.
      3. Make nice with Samsung
      4. Just pay the price to use a non-Samsung/non-Intel foundry and be at the mercy of someone else (again)

        • NeelyCam
        • 7 years ago

        You know your stuff. Welcome to TR!

        • anotherengineer
        • 7 years ago

        Obviously.

        I said, they have enough cash to and IF they wanted to. But who would when you have Samsung and others to take orders.

    • MadManOriginal
    • 7 years ago

    Oooor option 2: Apple and others move to Intel chips for their mobile devices 🙂 That’s the path Intel wants to take, obviously, and it’s not too far-fetched it just won’t happen ‘next quarter’. Just need software support…Apple has shown they’re willing to abandon microarchitectures wholesale if it means a better product, so why not iOS on x86? Android runs on x86 already, as does Windows 8. That’s all the major players right there.

    Boom, done.

      • raddude9
      • 7 years ago

      Oooor option 3: Intel starts making ARM chips again and sells the whole SOC to Apple. That way Apple gets to keep their binary comparability with existing CPUs and Intel isn’t reduced to just being a Fab facility.

      Boom, done slightly better. 🙂

        • Beelzebubba9
        • 7 years ago

        Intel won’t do ARM cores, though. And for good reason.

          • NeelyCam
          • 7 years ago

          And Intel has the upper hand in this negotiation. Intel doesn’t need Apple as much as Apple needs Intel.

          • raddude9
          • 7 years ago

          I can think of a number of reasons myself, but no good ones, please enlighten me.

            • Beelzebubba9
            • 7 years ago

            Intel has a huge competitive advantage in its process technology and a virtual monopoly on x86 CPUs in most verticals, so opening up their biggest advantage to an ISA that they can’t control or ever hope for any kind of vendor lock in with just doesn’t make sense.

            That and by this time next year I expect Intel to have SoCs that are much more competitive (Valley View), if not the best in the industry. Once 14nm hits they should have an even bigger lead and I wouldn’t be surprised if by 2015 or so there are a lot of x86 smartphones available on the market. .

      • NeelyCam
      • 7 years ago

      [quote<]Apple and others move to Intel chips for their mobile devices[/quote<] This is the eventual outcome.

      • TurtlePerson2
      • 7 years ago

      When Apple moved to Intel for the Mac chips they were in dire straits in terms of Macintosh sales. The move to Intel was a necessity and not really a choice on their part. They were never going to capture a large market when their hardware was completely different to develop for than x86.

        • TEAMSWITCHER
        • 7 years ago

        When they made the jump the deciding metric was performance/watt not instruction set. I would imaging that the iPad 4 enjoys a higher performance/watt than Intel chips right now. Intel is still a couple years away from a true system on a chip and forever on making it price competitive to Apple’s ARM chips.

        I have to believe that Intel is feeling some uncertainty. What Apple is doing with semi-conductors is unprecedented and could make companies like Intel, nvidia, and AMD obsolete. How do you compete against a hardware company that doesn’t have to buy their major silicon. If you can out perform them….great. If you can’t….then you’re screwed.

          • NeelyCam
          • 7 years ago

          [quote<]I would imaging that the iPad 4 enjoys a higher performance/watt than Intel chips right now.[/quote<] If you compare to Intel's 32nm, it's very close, but A6 is still leading. [quote<] Intel is still a couple years away from a true system on a chip and forever on making it price competitive to Apple's ARM chips.[/quote<] No - Intel already has true SoCs (the 32nm ones). As soon as 22nm chips land (in 2013), Intel has performance/watt and cost lead over Apple's chips (and everyone else's, for that matter)

          • MadManOriginal
          • 7 years ago

          It’s not unprecedented, it’s just more visible and perhaps on a larger scale because they are consumer products. If you know your semiconductor history going back to the ’90s you ought to be able to figure out why it’s not unprecedented.

    • brucethemoose
    • 7 years ago

    Come to think of it, Intel may be Apple’s perfect supplier. They have the capacity, they’re ahead of the competition for the near future, and they seem to dependably churn out die shrinks like clockwork, without any apparent hiccups. Intel can pick up the business they lose when Macs and other products go ARM, their processes are already moving towards a low power focus, and Apple’s demand for chips doesn’t wane as much as the rest of the industry. Apple, if anyone, is in a perfect position to absorb Intel’s higher prices, as ARM chips are cheap and product margins are already high.

    The big problem is that Apple is turning into Intel’s competitor… ARM is slowly eating x86 from the bottom up, so why would Intel choose to catalyze that? The other big problem: LP 22nm chips (the new atoms) aren’t out yet, and Intel didn’t even have that much LP 32nm capacity. Whatever process they use for broadwell won’t necessarily work for a chip that belongs in an iPhone.

    • Geistbar
    • 7 years ago

    If these analysts were any good at being analysts, they wouldn’t be telling the whole world the best idea going forward. Instead, they’d be being paid to tell other people what to do and then to never tell anyone else about it.

    The analysis seems to be based on an assumption that nothing Intel does will influence anything else in the market. It wouldn’t exactly be cheap for Intel to create sufficient capacity beyond their own needs at 14 nm. It would also remove a significant market advantage Intel has in having probably the best fab processes in the world. Plus the fact that it would be helping to improve an x86-free market and it could very well hurt Intel’s primary revenue stream. I don’t have access to all of the data, but I trust that the people making the decisions at Intel do, and I don’t see them chomping at the bit to do this.

      • JustAnEngineer
      • 7 years ago

      This sort of deal where you make an extra 28¢ today and lose half of your total business five years down the line is exactly the sort of short-term thinking at which analysts excel.

      • NeelyCam
      • 7 years ago

      [quote<]It wouldn't exactly be cheap for Intel to create sufficient capacity beyond their own needs at 14 nm[/quote<] I would say once the expensive process R&D is done, it would make sense to build as much capacity as the market demand can handle to profit more from those fixed up-front costs

        • BIF
        • 7 years ago

        Up to a point, that is. When all is said and done, chips are commodities; not unlike concrete, lumber, cold-rolled steel. The laws of supply and demand apply to Intel just as they do to Exxon-Mobil, US Steel or any large agro firm.

        Building and using too much capacity could alter the supply/demand ratios and would apply downward pressure on prices and margins. At some point, you’re not earning enough after taxes and expenses to pay for the R&D costs that got you there in the first place. Business is not a charity. Ostensibly, that is.

        • clone
        • 7 years ago

        Harley Davidson limited their motorcycle production capacity in NA to sub 200,000 for decades, by doing this Harley succeeded in maintaining demand which in turn strengthened their brand while also allowing them to dictate price and cushioning them from economic downturns….. Harley Davidson very foolishly doubled their capacity post 2000 and by 2007 the brand lost it’s exclusive cachet which limited it’s value, additionally because their product was a luxury item the recession of 2007 had a devastating effect on Harley which they are still struggling from.

        EXCLUSIVE TECHNOLOGY is Intel’s lynchpin, Intel maintains the best tech and leverages that position to dictate demand, their tech also buffers Intel if their products aren’t architecturally optimal, the analyst is likely hoping for better opportunities for Apple…. most definitely the stupid recommendation being parlayed is not good for Intel and can do nothing but hurt Intel in the long run.

        the all too common results of overcapacity in any market are shrinking margins and volatility as the costs to maintain a reasonable production level to maintain the facilities leads to a saturating of the market.

      • BIF
      • 7 years ago

      You are right on.

      The only thing I could see opposite of what you say would be if the Intel decision makers decided to reinvent Intel and phase out X86 as its primary revenue stream in favor of “something else”.

      Every decision made carries with it some risk of failure. Catastrophic failure, even. The same risks often apply equally to the “do nothing and hope for the best” decision. That list is long indeed. Steak and Ale, Bennigans, General Motors, Chrysler, Sears/K-Mart, just to name a few.

      • Bauxite
      • 7 years ago

      This x 1000

      Intel would have to be absurdly short sighted to get in bed with Apple like this.

      Apple wants complete vertical integration down to every bit of design and being a “dumb fab” for them is literally insane. They are skilled experts at keeping maximum margins for themselves and letting suppliers ride on the razors just to keep the “big orders” flowing in.

    • TurtlePerson2
    • 7 years ago

    How much control does a chip designer have when they use Intel Custom Foundry? From what I understand, they simply send Verilog-type code to Intel and get a chip back.

    For low power devices you want to use a different threshold voltage than high performance devices. This means you want a slightly different process. I don’t know if Apple’s low power chips would necessarily work with ICF.

      • NeelyCam
      • 7 years ago

      Intel just presented a paper in IEDM on their 22nm SoC process. They had multiple transistors – some for high performance, some for low power:

      [url<]http://www.eetimes.com/electronics-news/4403044/Intel-s-22-nm-trigate-SoC---how-low-can-you-leak--?pageNumber=0[/url<]

        • TurtlePerson2
        • 7 years ago

        It’s kind of funny that Intel’s chart explaining the versions of its new process more-or-less admits that the 22nm process is only a 30nm process…

          • NeelyCam
          • 7 years ago

          Yep. I guess they called it “22nm” because the 30nm FinFET performance somewhat matches extrapolated performance of a planar 22nm FET… It’s all marketing now anyways – the only way to figure out which fab has the best transistors is to wait for IEDM papers. Node names don’t tell much anymore

            • MadManOriginal
            • 7 years ago

            So it’s like PR numbers for process nodes? 🙂

            • NeelyCam
            • 7 years ago

            PRocess nodes

          • willmore
          • 7 years ago

          Is everyone else doing that for their chips, too, or is this just Intel? We hear TSMC and GloFlo talking about fin-FET processes and specifying dimension numbers to go along with them. Are these inflated values as well or are they more accurate?

            • MadManOriginal
            • 7 years ago

            They’re all inflated. Or deflated, depending on which way you want to look at it 😉

            • willmore
            • 7 years ago

            I’ll believe you when I see a cross section SEM image from Chipworks! Until then, death to the infadel!

            • NeelyCam
            • 7 years ago

            TSMC’s 28nm:

            [quote<]"Contacted gate pitch is ~118 nm in our initial analysis, with minimum [b<]gate length of ~33 nm[/b<],"[/quote<] [url<]http://www.chipworks.com/blog/technologyblog/2011/07/12/more-hkmg-hits-the-market-gate-first-and-gate-last/[/url<]

            • willmore
            • 7 years ago

            Well, that settles it.

          • JumpingJack
          • 7 years ago

          Where did you get that Intel’s 22 nm process is actually 30 nm??

            • NeelyCam
            • 7 years ago

            [url<]http://www.eetimes.com/electronics-news/4403044/Intel-s-22-nm-trigate-SoC---how-low-can-you-leak--?pageNumber=0[/url<] Check "Lgate" (=gate length) on the table. The smallest one is 30nm

    • Decelerate
    • 7 years ago

    It would be a logical move if those chips are sent to the US manufacturing site. Otherwise it’s closer to ship from Korea/Taiwan to the assembly plants.

    However Intel would have to commit to some medium term contract at least. I have 3 years in mind if they’re producing smartphone logic, as Apple tends to keep its product around that long, based on the 3G & 3GS retail strategy.

      • ludi
      • 7 years ago

      Numerous chips are packaged far away from the foundry site. For example, both Intel and AMD have final CPU packaging facilities in Malaysia even though most of their chips are fabricated in the US and Germany. Shipping costs have relatively little to do with it, the logistics are already in place and both the intermediate and final products are not particularly heavy or large on a per-unit basis.

        • Decelerate
        • 7 years ago

        Interesting, thanks for the info. I saw this as such an… unoptimal process.

          • ludi
          • 7 years ago

          Chip fabrication, particularly on leading-edge foundry processes, is a capital-intensive and high-skill process. Final packaging and testing is less capital intensive and requires less skills. So, as long as the shipping costs are relatively low, it makes sense to geographically divide the facilities based on competitive advantage.

          Also, a company the size of Intel also tends to rotate their product and fab upgrades in order to maximize the life of the equipment — this year’s leading-edge fab will make CPUs, but for the next year or two it makes chipset components and for another year or two after that it might be making memory products or other peripheral components. After that, it might be next in line for a massive capital upgrade to the newest process, and back to making CPUs again. So, at some point the fab might be making products that could theoretically be manufactured more cheaply by a SE Asian facility, but since the equipment has already been procured and amortized to support higher-margin lines, the remaining production life is basically gravy.

        • NeelyCam
        • 7 years ago

        Intel/AMD might consider building assembly capacity inside the USA, though… helps with political landscape (“American jobs!!”) and ITC complaints

    • Silus
    • 7 years ago

    Apple wants the cheapest chip prices, something that Smasung has provided but has had enough, especially since Apple behaves as the spoiled brat that it is when it comes to useless patents.
    Intel won’t be cheap and Apple will need to pay to the nose in order to use their fabs. But that’s what they get from wanting everything their way or the highway.

      • brute
      • 7 years ago

      any patent that can be used to remove a competitor from a market or gain a billion $$ settlement is far from useless

        • Silus
        • 7 years ago

        When patents are granted by monkeys (from the USPTO) that don’t even check for prior art…yes, they are useless.

          • MadManOriginal
          • 7 years ago

          The reason ‘monkeys don’t even check for prior art’ is because that’s the way the USPTO is meant to work now. It’s supposed to save taxpayer money and make the process faster, any validity disputes are supposed to go through the courts. It seems to me that the intent of saving taxpayer money isn’t being met though, it just moves the burden over to the court system. It was probably some lawyers who thought it would be a good idea to do it this way 😛

            • Silus
            • 7 years ago

            Precisely and that excuse makes no sense because of what you just said, since I’m pretty sure that long battles in court, plus appeals and whatnot cost a LOT more to taxpayers than a “monkey” (or several of them) actually doing their job of checking all that is necessary before granting a patent!!!

            • ludi
            • 7 years ago

            You can use lots of exclamation points!!! but the underlying change needs to come from Congress. Software and business process patents, in particular, opened Pandora’s Box. The PTO is just doing the best they can with the constraints they are given.

            Also, such reform might be surprisingly expensive, because people who have the necessary smarts to properly check all of the stuff that presently gets crammed through the PTO usually have much better opportunities in private industry — i.e., you’ll have to entice them with commensurate salaries.

      • Beelzebubba9
      • 7 years ago

      Apple is certainly willing to pay top dollar for components or services when there’s a sound engineering or design reason to do so. No one in their right mind can look at the displays Apple user in the rMBPs and dribble out ‘lol cost cutting’ in response. Or on their phones. Or the fact that they designed their own CPU core in house for the A6.

      I think Intel would be very stupid to fab chips for Apple, but it’d certainly make a ton of sense from Apple’s perspective even if it drives up their per-unit SoC cost substantially.

    • chuckula
    • 7 years ago

    1. TSMC’s “14 nm” process isn’t really 14 nm.
    2. An early 2013 tape-out at TSMC means you can expect to be able to buy products made using the process in early 2015… if everything goes perfectly. Ask Nvidia or AMD about whether things always go perfectly with the TSMC processes.

      • brucethemoose
      • 7 years ago

      Meanwhile, Intel will probably have 10nm skymont ultrabooks on the shelves by the time TSMC’s 14nm process is sorted out.

      It’s not that TSMC is bad… it’s just that against all odds (and the laws of physics), Intel keeps churning out die shrinks on schedule.

      • TurtlePerson2
      • 7 years ago

      What do you mean by “isn’t really 14nm”? Is the process not capable of 14 nm gate lengths?

        • MadManOriginal
        • 7 years ago

        Some of the component gate lengths are ‘really’ 14nm, and some are 20nm. I think it’s a common thing for all foundries at various nodes, it’s just not talked about much.

        • NeelyCam
        • 7 years ago

        Going from 20nm to “14nm”, TSMC didn’t scale down contacted gate pitch or metal stack. Even if the gate length is in fact 14nm (it might not be – the “22nm”, “20nm”, “14nm” etc. monikers are mostly just marketing labels now… I think it was TSMC that pointed out that Intel’s 22nm isn’t really 22nm but closer to 30nm), chip sizes won’t scale much because they can’t pack transistors any denser than in 20nm.

          • TurtlePerson2
          • 7 years ago

          But with the W/L term in Ids, you still get better performance from your transistors even if you can’t pack them tighter.

            • NeelyCam
            • 7 years ago

            Yes, and because of that improved performance some minor area scaling might be possible. But if the process gets substantially more expensive per mm^2 (likely with the inclusion of “14nm” FinFETs), chip size scaling is pretty important to keep the chip cost down.

            Then again, maybe performance (or, rather, performance/watt) is so important than even with a price increase the chips would sell.. of course, depending on what the competition (GloFo/UMC/Samsung/Intel) is doing

      • Helmore
      • 7 years ago

      I think you’re talking about their 16 nm process, in which case you’d be right. Are they even planning on a 14 nm process?

        • NeelyCam
        • 7 years ago

        In reality, anything marketing calls something between 11nm and 19nm is pretty much the same thing

Pin It on Pinterest

Share This