Samsung’s Future is in Competing with Intel not Apple

Now that many major silicon fabs are embarking on 14nm, things are starting to get very interesting with relation to the future of the semiconductor industry. Both Intel and Samsung are pushing forward on the 14nm process technology while others are struggling to get to 14nm with a quality process. In light of this, I have a few new observations about the future of the semiconductor industry.

A new dynamic has come into play about future nanometer process nodes and whether or not certain companies can achieve greater than 14nm, 10nm, and beyond. Intel is going to get to 10nm and there is a high likelihood they get to 7nm. Whether Intel can get to 5nm we will have to wait and see. Samsung is on 14nm now with a very high quality process and this is why companies like Apple and others are looking to Samsung to make their chips using Samsung’s 14nm process technology. Interestingly, Global Foundries has entered into an agreement to license Samsung’s 14nm process technology and use their manufacturing capacity to make chips on Samsung’s 14nm FinFET process for their customers. Making things even more interesting, there are reports Qualcomm will also look to use Samsung’s 14nm process technology for a line of Snapdragon processors.

What this is signalling is the difficulty fabs like Global Foundries and TSMC have had in getting to 14nm with a quality and reliable process technology and how important getting to future technologies is in general. Samsung seems to understand this as they are reportedly building a massive facility to manufacture chips. This is relevant, and telling, because it indicates Samsung may be looking to take on Intel rather than Apple in the future.

Most people know Intel makes chips, but they may or may not know they have a proprietary architecture called x86. This architecture is different from the ARM architecture which Qualcomm, Samsung, Apple, and others use. Intel’s belief is they may be the last company standing when it comes to cutting edge process technology. There may be some truth to this. David Kanter is a well know semiconductor analyst and he had a very insightful tweet last week.

Screen Shot 2015-05-11 at 2.58.55 PM

What he is insinuating is the difficulty getting to new process nodes is proving to be a stumbling block and thus narrowing the field. Using this logic we can ask an interesting question — what if Intel is the only company who can get to 7nm or even 5nm? Will every other fab then have to license Intel’s process to make cutting edge future generation chips? I ask this question to make a point. It seems we can make an argument that not everyone is going to get past 10nm and therefore, if only a few do it, options are limited. Perhaps one, two, and possibly only three companies will bear the load of making chips for the whole of the industry. Whoever is in this position has quite the upside. I believe Intel will get there and Samsung appears to be gearing up to take them on. However, this winnowing of the field when it comes to process technology is a fascinating way to look at who may be relevant to making semiconductors in a few years.

Samsung seems willing to license their process technology to other foundries and the question is whether Intel will or will not some day. Assuming Intel is the last company standing when it comes to 5nm or beyond, they don’t have enough physical space to make everyone’s chips. It is likely licensing their process technology is a viable option. The other fascinating element to the Intel story is whether they will license their process out for others to make ARM chips. I don’t believe they should force x86 on the world if their process technology is the last man standing, which is why, by the time Intel is at 10nm and certainly by 7nm, if they are not making ARM chips for someone on their process or have licensed their process out for ARM chips, we are going to need to start asking very difficult questions of the company.

Ultimately, the super bear case for Intel is that they do not or can not maintain a process technology lead. Which is why this article stating Samsung has already demoed a 10nm FinFET chip ahead of Intel is interesting at a high level.

How this plays out will be fascinating but, in the case of Intel, Samsung, and others, my focus is on their ability to manufacture at future process nodes or not. This, not architectural debates, seems where the real competition for the future of the semiconductor industry lies.

Published by

Ben Bajarin

Ben Bajarin is a Principal Analyst and the head of primary research at Creative Strategies, Inc - An industry analysis, market intelligence and research firm located in Silicon Valley. His primary focus is consumer technology and market trend research and he is responsible for studying over 30 countries. Full Bio

29 thoughts on “Samsung’s Future is in Competing with Intel not Apple”

  1. Isn’t there a contradiction between touting Apple’s very limited in-house R&D and half-baked vertical integration, and arguing Samsung should let go of that and give up on making finished product ?
    Even if you see the finished product as simply a conduit for parts (ie, discount any financial gain at the finished product stage, though Samsung are still making billions at it even in their off years), has any other OEM shown an ability superior to Samsung’s at realizing that “parts value” ?
    The 3 or 4 stages of the value chain (parts, OEM/ODM, marketing/distribution/branding, ecosystem building –optional–) are only loosely coupled (plenty of companies are living sometimes thriving on only some of those) and involve different skillsets. Conglomerates are supposed to be able to excel at each stage, and provide some synergy/vertical integration (those words vaguely nauseate me though… shades of Jack Donaghy).
    I don’t think the fat lady has sung yet, and looking at the current state of affairs, Samsung’s parts business doesn’t seem to be suffering from working a lot for Samsung’s OEM side (it’s capacity-constrained anyway, why not capture the extra margin in the finished product ?), and the OEM business benefits mightily from being able to play it both ways, sourcing either internally (avoiding this year’s 810 fiasco, getting top-of-class screens cameras CPUs RAM) or externally when needed (capacity, cost…).
    We keep hearing about differentiation… I don’t think the software/services side is the only one, hardware differentiation still exists, can contribute to success, and doesn’t *have* to be a distraction.

    1. Yes I’m not saying the OEM business goes away. But only using the OEM business as the primary shell to move components is not going to grow them the way they need to grow.

      It is necessary for them to start selling more components more broadly and anyone who understands the chipset side knows how profitable this can be if they are a major player.

    2. “touting Apple’s very limited in-house R&D and half-baked vertical integration…”

      Such nice, compact and concise trolling. Your mother must be proud.

      1. Could be worse, I could just comment on others’ posts, middle-school style calling on their mothers ?

          1. Re: “Say hi to your mother for me okay? :)”

            My, oh my. Should I point you to redditforkids.com ? That kind of drivel has nothing to do here.

          2. You don’t know the “Say hi to your mother” Mark Wahlberg bit? Come on dude, it’s hilarious, and relevant, you set it up perfectly. I had to do it. The Gods of Comedy demanded it!

          3. Sorry, not a US person. And I’m not your dude, far from it. Go count your Apple stock somewhere else, and to a comedy site, you’ll feel more at home for sure.

    3. Samsung’s OEM phone business is being squeezed mightily from above and below (S6 unsubstantiated sales bluster notwithstanding) so the point is that non-OEM is worth pursuing heavily since they have at least process advantages (even if their designs are lackluster in performance per cycle/core). Remember, the Samsung components business treats its OEM business as another customer so there is little risk in pushing for more customers for their fabs.
      Samsung doesn’t really control the user experience with their own phones and where they have tried to take charge of it, it hasn’t been particularly successful (Touchwiz and myriad poor S-Apps) and they have dialed it back with the S6 (a small minority of total OEM sales). Apple’s chip designs are clearly better than Samsung’s in power and efficiency and their spending on manufacturing capital ($billions per quarter) embarrasses most “pure” manufacturers. The fact that the dedicated machines go into Foxconn factories absolutely creates meaningful vertical integration. I would say that Samsung is the one that has a real lack of the top-to-bottom stack… manufacturing is increasingly a commodity these days. Apple’s version is (as usual) much stronger than you give them credit for.

        1. That Samsung needs 8 cores to take on Apple’s 2 and that they need to cheat in the benchmarks to compete? Samsung’s competing chips also come out 6 months after Apple’s.

          Was that was the main point you got from my piece? Not the fact that Samsung is NOT vertically integrated regardless of the fact that they own some factories… ho hum.

          1. Again, what’s your source ? You say Sam’s chip are lackluster per-core, my data says they’re the same or slightly ahead… Do you have any source at all, or is your rant just made-up (which I’ll assume it is, unless you come up with the goods, throwing a shadow on all the rest of your post) ?

            Also, the “Samsung is not vertically integrated” part was so out there I just disregarded it. FYI, Foxconn do not make chips for Apple, Apple actually mostly buy their chips from Samsung…

  2. I don’t care much about these competitive business matters, but I can’t help but ask.

    How does the technological advancement of silicon fabrication rule out competing for devices as the title states? If anything, Samsung is far more independent than Apple in producing their own devices. They make cpus, ram, screens, flash ram, etc.

    Might Apple need to build their own fab is a more interesting question. Samsung devices are not going to just go away. That strikes me as wishful thinking.

    1. Yes I’m not saying their end brand units go away. Only that it is inevitable they begin selling SoCs to other vendors to capture growth in an area they need it.

      Over the past few years the component business has grown to do more revenue than their OEM business. I don’t see this changing, and even if they do remain steady in handset sales it will continue to be in low-margin areas. Samsung’s growth in revenue and profits has more upside in this area and more component side than OEM. It is a shift for them and that is my point in where their focus should die.

  3. How does this change if Intel and the rest can’t go beyond 7 nm for example? If 7nm is the practical limit for the next 5 years, doesn’t that give every other foundry time to catch up to Intel? Intel lives and dies by Moore’s law. If that stalls or goes away, Intel likely loses whatever lead they have in a very short time.

    1. So the issue is in whether or not the other guys CAN catch up. The more I speak with those who understand the technical bits of this much better than myself, all emphasize just how difficult getting beyond 14nm is and will be. 10nm is in sight but there are some extremely difficult problems to solve getting beyond 10 and 7nm which are the reasons the debate over Moore’s Law ending are wrapped up in.

      From what I hear from the experts in foundry process it is going to be hard and hence the point that others may not catch up. Leaving us left with only a few who have the process and can license and make for everyone else.

      1. What is hard right now probably will become easier once the processes are perfected. STMicro is leaving because it isn’t economical for them to compete on creating smaller nodes. But if the process stalls for a period of time, EUV becomes a reality, and the costs associated with manufacturing become well understood, then old foundries or new can reevaluate the market and make a decision.

        I think this only works if there is enough time between node shrinks. It is a situation that hasn’t occurred before so I’m speculating but a stall in Moore’s law changes the economics of semi manufacturing and I don’t think the market will stand for a single manufacturer.

  4. Interesting article Ben.

    Fascinating to think of the ramifications of only one company dominating the best fab technology. Even if only two companies make it, the industry would look quite different to a single dominant company.

    Bizarre that Intel hasn’t started Fabbing Arm chips yet – how long will it take for them to acknowledge their mobile strategy isn’t working?

    1. That is the question. I’m also not sure, even though they should, that they end up making the ARM chips. I can see them license their process first to others who will make it in their foundry. I think there are more open doors there now than ever, it is just them figuring out the right balance and model.

  5. i hear you guys arguing this all the time, but what i find odd is you just don’t discuss what is obliviously public knowledge and tackle the cost vs benefit and the actual technical hurdles. Ie can intel even get to 7nm economically without more volume outside of x86 or captive customers like altera whom it was rumoured to acquire? aren’t there 3 fabs that get to 10nm guarenteed already , not including GF who share tech with samsung at 14nm? what is 10nm anyway? intel claims their nanometer measurments at 14nm is more dense, true scaling, not just the fins interconnects or whatever? …beyond all that, if we get to 7nm, what is the cost, efficiency benefits theoretically vs 10nm … is 30% reduction in power and performance, where does the efficiency and cost benefits of scaling stop even if u have unlimted capital …5nm, 2 or 3 ?

    1. So I’d add another way to look at this. While you are exactly right about the economics of this, it does emphasize my point. Intel is insinuating they can see the economic benefits extending to THEM up to 7nm. Their ability to get there with the science and economics is just another point to who may actually get there. I’ve heard the statement that economics will kill Moore’s law behind the science and that may be true. But if only a few companies can not only accomplish this technically and economically then it adds to my point about whose process will be the last man standing.

  6. An equally fascinating question for me is whether Intel’s goal is to be the sole foundry capable of going to the smallest process, and using that as a way to make x86 relevant in the low power consumption game. While licensing out fabrication seems to be the natural thing to do, they could just as easily choose to use their advantage in fabrication as a moat for x86, and then a spiked wall against ARM (I do not like this scenario, but it seems plausible to me).

Leave a Reply

Your email address will not be published. Required fields are marked *