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!

  • Intel to Skip 10nm to Stay Ahead of TSMC and Samsung?

    Demystifying Analog & Mixed-Signal ASICs-intel-itanium.jpgQuarterly earning calls are a great source of information but they can also be a source of confusion and generally it is an unhealthy combination of the two. On one hand these earning calls are to appease the financial community. On the other hand, in my opinion, these calls are also used to generate fear, uncertainty, and doubt amongst the competition, absolutely.

    Intel's (INTC) CEO Brian Krzanich on Q2 2015 Results - Earnings Call Transcript

    Foundry Business
    Before we get to the prepared statement I just wanted to point out that Intel Custom Foundry (ICF) was not mentioned in the call nor was it part of the Q&A. Intel (and most publicly traded companies) generally do not announce when they are quitting something (Itanium for example), they just let it go quietly off into the sunset. My guess is that Intel will not continue in the foundry business.

    Today the foundry business really does revolve around mobile SoCs. It started with Qualcomm at 28nm but Apple took over at 20nm and again at 14nm. By revolve around I literally mean that the semiconductor manufacturing processes are built for SoCs and adapted for everything else. Apple writes some REALLY big checks and that is what makes the capital intensive foundry business work. Apple will again “influence” 10nm and 7nm with their technical and financial prowess and that just does not fit with Intel’s process development culture, my opinion.

    Mobile Business

    “We have also updated our mobile roadmap. Our OEMs' first Atom x3, x5, and x7 products were announced and are ramping using our previously code named Cherry Trail SoFIA 3G and SoFIA 3G-R products. The 4G version of our Atom x3 platform, SoFIA LTE, is sampling now for network certification, and is expected to ship in volume in the first half of next year. Our latest LTE modem, the CAT-10 7360, is on track for shipments to customers this year.”

    Again, I doubt Intel will ever announce that they are quitting mobile but this is a strong signal that they are scaling back. I have also been told that Intel is cutting mobile staff and shutting down complete mobile groups all over the world.

    At the top end of mobile you have device makers Apple and Samsung (who make their own SoCs) which control the majority of market share and profits. Then you have SoC giants Qualcomm and MediaTek who control the majority of the rest of the mobile sockets leaving Intel and the Chinese SoC companies fighting unprofitably for the final few. So yes, Intel will quit mobile at some point in time, my opinion.

    Altera FPGA
    “I'd like to shift gears now and talk about a couple of important strategic updates. Last month, we announced our plan to acquire Altera, a leading FPGA vendor. We see four key strategic drivers behind this acquisition. First, we believe we can enhance Altera's base FPGA ARM-based business substantially. We plan to do this through our leadership in Moore's Law and our ability to execute designs using our tools and silicon more quickly, allowing us to continue to support and develop their ARM-based products."

    I really don’t understand the “ARM-based products support” statement. If you were designing an FPGA with an ARM core inside would you really chose Altera/Intel? FPGA designs have a VERY long shelf life and I would not bet my career on the remote possibility of a healthy relationship between Intel and ARM. Not today anyway. ARM is making another play for Intel's Data Center business, right?

    "Second, history tells us that the FPGA vendor who is first to a manufacturing process node enjoys a market segment share advantage over the life of that node.”

    Yes the FPGA vendor that hits the new process node first is awarded extra market share. Altera beat Xilinx to 40nm by a year or more (depending on whom you ask) and dominated that node. Xilinx came back and beat Altera to 28nm by a couple months and reclaimed leadership. Xilinx again beat Altera to 20nm by a significant margin and will dominate that node. At 14/16nm both Xilinx and Altera taped-out last quarter so it is too close to call but my bet is on Xilinx and TSMC 16FF+. It really is an amazing process and will ramp quickly. 10nm will again be close but as we have read Intel has delayed 10nm and Xilinx is already working with TSMC on 7nm so the smart money is on Xilinx.

    Intel 10nm Update

    "The last thing I'd like to share with you is an update related to our 10-nanometer technology transition..."

    BK did a nice job of spinning the 10nm delay in the prepared statement. In the Q&A however there was a more direct question and response:

    "No, I'd call it similar to what happened on 14-nanometer. Remember, on all of these technologies, each one has its own recipe of complexity and difficulty, 14-nanometer to 10-nanometer same thing that happened from 22-nanometer to 14-nanometer."

    One of the possible scenarios I see here is that Intel will improve the performance of 14nm (similar to what TSMC did with 16nm and 16FF+) and skip 10nm in favor of accelerating 7nm. This makes complete sense if Intel wants to maintain their process lead against TSMC and Samsung. It also makes sense if Intel wants to continue to cut expenses. Sound reasonable?

    Everything Else

    There was also a lot of good news in there about the PC, Data Center, and NAND business in which they are dominating. The Intel IoT business is of interest to me but I don’t really understand it. This is the old embedded group, right? What growth path was it on before they renamed it IoT? And where exactly is that 4% growth coming from?

    You can find the full transcript HERE.

    Also read: TSMC (Apple) Update Q2 2015!