Why ASICs are Important (PA Semi & Apple )

There are few product platforms that can support the development of unique processor or ASIC. Apple is one of those companies that has product platforms that drive enormous volumes. For semi developers this shrinking list of prospects is a nightmare – you win apple or you don’t. If you lose, you have nothing. Thus there are fewer application specific devices getting developed by semi-vendors these days.

This feels like Cisco’s Quantum Flow device thought process. I’ll speculate that it goes like this “We are the leader – we are the market – it’s all about software”. Hmm what hardware are we going to use? There is no processor being developed that fits our needs. We’ll have to sub it, but the sub only has us as a customer. We’ll have to show them what to do. Then the sub will try to sell this to our competitors who will have an equivalent hardware platform to develop software on. Yuck. Let’s do it ourselves. Bottom line — these monster platforms can be vertically integrated if they want to.

Links

Relevant NPU & SoC Posts

  • System-on-a-Chip (SoC) vs Network Processor (NPU) – What’s the Difference?
  • NPU Positioning – Where do they win?
  • Cisco hasn’t given up on hardware … New Quantum Flow NPU
  • Advertisements

5 Responses to Why ASICs are Important (PA Semi & Apple )

  1. … and the thing is… they don’t care if it costs more than buying off the shelf either because of the intangible benefits it brings. NPU vendors and others who tout the cost advantage of their products are missing the value the customer cares about.

  2. Iain Verigin says:

    The cost advantage is a complicated beast.
    Cost has two major parameters – 1) component cost and 2) development cost.

    For a “monster” platform ASIC there is only one 1 customer. This customer usually does a lot of ASICs so it’s component cost will be very competitive with a
    “specialty” semi vendor. The place where the customer can save big is on development cost. For example Quantum Flow was reported at $100M. This is only going up.

    For the majority of platforms. ( think …. everything but a handful) the devices can be sold across lots of platforms. In this case the semi-vendor has much higher volumes than any single “box” vendor and can leverage development cost across many programs.

  3. DinoP says:

    One more criteria to be taken into account.. for Apple to want to get into their own ASICs in the consumer space, this must be a generation product. That’s to say, a processor that will power a family or generation of products. If they thought they’d see additional or changing requirements every 6-8 months they might be best served with the newest processor in the space but oh ya not to many of those out there either.

    This generation product will most likely cost more due to number of devices it may be trying to serve in the family or over the generation.

  4. Iain Verigin says:

    Thanks Dino. Good point. Yes it could become a family of devices.

    fyi … Andrew noted in an e-mail to me that there is also the “scorched earth” aspect of doing it yourself.
    – in my words — If you make the semi market TAM for a segment “bad”, by removing a significant chunk of it, then there will be less “semi” investment in that space. This makes your segment more defensible. This of course only works when one has extremely significant market share in a segment. Which is why it works for Cisco’s QFP and maybe Apples consumer devices.

  5. […] differentiator left for CPU designers would be Application Specific Features. It’s those darn ASICs […]

%d bloggers like this: