You are currently viewing SemiWiki as a guest which gives you limited access to the site. To view blog comments and experience other SemiWiki features you must be a registered member. Registration is fast, simple, and absolutely free so please, join our community today!

  • Further Delays for Intel 10nm?

    Intel's 10nm may be reliving the 14nm elongated delay issue-
    Schedules & tool delivery may be pushed even deeper into 2016-
    Meanwhile Samsung & TSMC press on-
    Could Intel be embarrassed?

    Article: Introduction to FinFET technology Part II-intel_tech_lead.jpg

    Deja Vu---14nm all over again?
    We are hearing from a number of people in the industry that 10nm development has had similar yield ramp issues as Intel experienced at 14nm. Although we have been among the first to report about the delays a while ago we are hearing about further potential delays beyond our first reports. We had originally anticipated some significant tool move ins and an acceleration of effort by Intel in the second half of 2015 which was pushed into 2016. We are now hearing that more of the ramp will be pushed even deeper into 2016. It is unclear what the issues are but obviously new nodes are getting harder and harder as process complexity goes up exponentially. The delay may also not be 100% technical and may be a choice to slow spending as well.

    This is obviously a critical issue for Intel as their primary value in the industry has been maintaining technical leadership & dominance. If that technical dominance is threatened so too is the value of the company and its place in the industry threatened. We think Intel's technology lead is no longer a given and 10nm may be a highly critical flash point whose outcome will determine the future success of many of Intel's products.

    Intel maintaining radio silence on 10nm..
    Given that the public handling of the 14nm roll out and issues was less than optimal, Intel has kept a tight lid on comments regarding 10nm. While we agree with this policy as its best not to say anything until the situation is well under control it does open us up to a lot of speculation and comparison as to where other chip makers, most notably TSMC and Samsung, are in their efforts versus Intel.

    Like it or not, there is a lot of public posturing between the three companies which impacts customers, investors, suppliers, employees etc... While not opening your mouth prevents foot insertion it allows others to drive the messaging in the near term. Samsung and TSMC have been happy to oblige.

    Samsung 10nm Exynos & 14nm A9
    Back in February at the 2015 International Solid-State Circuit Conference (ISSCC) in San Francisco, Samsung announced the first 10nm device. What that really meant is very hard to decipher and separate out the marketing BS from the technical reality but the gauntlet was clearly thrown at Intel's feet while Intel had duct tape over its mouth. What we have heard since then is that Samsung is pushing to have its 10nm Exynos in its smart phones by the middle of 2016 (about a year from now). If they had first 10nm silicon in February that is not a totally unreachable schedule and Samsung could also force the issue by producing a bazillion wafers to compensate for low yields if they can't get it up. This would obviously be a marketing coup for Sammy as they could have a phone with a newer processor out before Apple's standard fall 2016 release schedule. Meanwhile Samsung is busily producing Apple A9 parts for this fall's rollout for Apple, while still producing Exynos parts at lower yields.

    This "Frenemy" concept is well beyond confusing.......
    Could TSMC be getting the A10? Even though A9 parts aren't even soldered into iPhones yet we have to start speculating about the A10. At this point we would lay odds on TSMC for a number of reasons. The Frenemy competition is heating up again as Sammy is pushing very hard to "scoop" Apple at 10nm. Apple has to keep TSMC in the mix to keep competition alive and thus continue to get the best prices and service. Lastly, TSMC has gone whole hog to get to 10nm and may have already elicited a promise from Apple.Time will tell, but right now that's how I would bet.

    Unfair comparing Apples & Oranges...
    The reality is that not all 10nm are created equal and Intel 10nm parts are still higher performing than both Samsung and TSMC parts if you look at the specifications. If you look at the alleged gate pitch, metal pitch & Fin pitch (pitch is spacing dimension) & SRAM cell size of Intel versus others, Intel is clearly superior. The problem is that 99.999% of the population doesn't know and doesn't care. In addition, its been less than clear that the dimensional advantage that Intel has has actually translated into performance advantages in real life. We had commented early on about the fact that the A8 and Broadwell were not all that far apart in specsmanship despite being manufactured at two different nodes with and without FinFET. However, at the end of the day the only thing that investors and consumers will hear is that Sammy beat Intel at 10nm and the rest will be incomprehensible background noise.....

    Also read:
    Who Will Lead at 10nm?

    Collateral Impact...

    At this point we don't see much negative collateral impact if Intel shifts 10nm further to the right. We have been very clear about delays for a while and they are built in to all the semiconductor equipment company expectations so there is no real surprise potential as expectations are already low. 2015 10nm spending has a low expectation. Intel could cut another billion or two off of already reduced numbers which would make overall WFE spending growth numbers more difficult to achieve but wouldn't impact 2016 unless the delay got so bad as to push spending out of 2016 into 2017. Its also unclear where 10nm will wind up. Talk has been of 10nm being done in Israel but Intel still has an empty fab in the US. We haven't heard of any movement in either location.

    The death of Intel mobile aspirations...
    If Samsung makes good on its 10nm Exynos mobile chip in mid 2016 and TSMC pushes ahead with Qualcomm and maybe Apple, where does that leave Intel's plan of being a competitive force in the mobile/tablet space? DOA.

    Intel has to worry about where its bread is buttered and right now that's the server market. New technology roll out will be scheduled to protect the sacred cow of servers and mobile/tablet will be an afterthought. That is not going to fly very well when the mobile/tablet market is the laser like focus of Apple, Sammy, TSMC & Qualcomm. The problem is that the laptop market is next to be threatened, especially by the A10 as the technology will migrate upstream from mobile/tablet market. Given that the 14nm roll out schedule has been long and slow we don't see a major change in velocity for 10nm and that will give competitors an advantage as their priorities are the reverse order of Intel's, mobile first others later....


    What ever happened to being a foundry?

    Aside from the mobile/tablet market the next obvious question for Intel is what happened to being a foundry player? If you don't have a viable technology advantage to offer customers to lure them into working with a difficult, foundry newbie, then what do you have? Certainly not lower pricing. Both the results and relationship with Altera on the foundry front has been less than stellar. Obviously buying them will keep them a captive customer rather than see them go to TSMC as some had speculated but shotgun weddings rarely have a lot of love and success. With no 14nm Altera silicon in sight (let alone taped out...) we have heard that Xlinx couldn't be happier about competing with a constrained Altera. If you love something set it free.......

    While it may seem totally stupid at first blush it may be an interesting exercise to have the Altera division of Intel free to choose between having its product foundered by TSMC or Intel and see what they choose as TSMC may have the process available first (what a scary thought...). Having a captive customer creates complacency. Another tidbit that raises our concern about the foundry business is Intel pushed out the VP of their foundry business with no replacement in sight......not a ringing endorsement.

    Other Changes at Intel

    There have been some management changes at Intel over the last few months, on the technology side, perhaps to shake things up and get back on track which we think is a good thing as you can't fix a lot by doing things the same way if its not working.

    We might take the opposite tack in terms of spending however. Rather than try to manage the numbers down in the near term by cutting spending to keep Wall Street happy with EPS, Intel may be better served over the longer run by doubling down and going all out to preserve its technology heritage. While there is clearly some clean up that can be done at any large organization that has grown over many years we would take the opposite approach and be pushing even harder on the core R&D and manufacturing side that got Intel to where it has been and remains the crown jewels of the company. Spending money on maintaining the company's crown jewels is far more important than any acquisition.

    TSMC and Samsung are not backing off the accelerator why should Intel? 10nm is a three car race of "chicken" with only enough room for one car to exit at the end of the drag strip and Intel is backing off the gas. It may be the conservative response, but not the winning response and there is a lot at stake in this race.

    "There is at least one point in the history of any company when you have to change dramatically to rise to the next level of performance. Miss that moment-and you start to decline." Andy Grove

    Deja Vu all over again...


    Robert Maire
    Semiconductor Advisors LLC