Foundry will never be more than a fun and modestly useful sideline for Intel.
Intel believes it needs to own the product to win the market. It wants to define the system (aka platform) and sell several high value chips into it. To succeed, it needs to be at the leading edge of process technology, but it doesnít really want to share that technology in any significant with anyone else.
It certainly wonít share its technology with anyone making computer chips. These days that includes mobile application processors for smartphones and tablets that are the new personal computers.
Intel wants the smartphone and tablet to use an Atom x86. Qualcomm and other ARM licensees could be on the verge of expiring for lack of the latest process technology and Intel would not turn on the tap for them.
Intel is dabbling in the foundry business as a fun sideline. It makes a little investment in an orthogonal market like FPGAs (Achronix, Tablula) or network processors (Netronome), then it gives them a little shove forward by giving them a taste of its 22 nm tri-gate technology.
The startups donít need many wafers. They take a few dollars from companies such as Altera, Broadcom, Cavium or Xilinx and put them in Intel Capitalís pocket. In the course of this side business, the startups may help Intel refine its latest process.
But donít expect Intel to open the flood gates on the foundry business anytime soon. Intel wants to dominate the world with x86 chips, not wafers.
If you really think about it, why would Intel bother with Foundry business which has a relatively modest profit margin that would not support Intel's independent R&D effort and would actually be a "self-goal" that would reduce it to the level of IDMs who could not keep up. So INTC is not just going after the Foundries who build the SoCs for Smartphones but the Fabless chip design companies right here in the US who have been maintaining a 30 % + profit margin. This requires convincing the system builders to switch from ARM to X86 once Windows 8 comes out ( revive the old Wintel duopoly of PC days ? in this case Android would be like CP-M and xxx like Zilog Z80 that used to dominate before 8086 ).
Yeah, it was supposed to be a big deal to design with Atom cores on TSMC processes.
I guess even Qualcomm when hard pressed for 28nm capacity for Snapdragon would still rather help others bring up 28nm processors than switch to Atom-based designs.
I agree with what Dylan mentions. Foundry gives Intel the edge towards pushing Atom. Consumer demand is there and if ARM based devices cannot feed in to this demand due to lack of foundry support then Intel gets what it wants.
I remember some years ago, Intel ported some Atom-related libraries to TSMC. What is the status of the transfered information? Is it practically useless?
It reminds me having Intel as a foundry may make it harder to switch other foundries, due to more restricted control on the libraries from Intel, being derived essentially from Intel's competitive process technology. I am sure most foundry customers still want this freedom.
Is there room for Intel to use its foundry muscle as a strategic initiative to push Atom? In other words, what if Intel said: Use Atom for your applications processors and we will build your chips for you (for a small fee, of course)?
David Patterson, known for his pioneering research that led to RAID, clusters and more, is part of a team at UC Berkeley that recently made its RISC-V processor architecture an open source hardware offering. We talk with Patterson and one of his colleagues behind the effort about the opportunities they see, what new kinds of designs they hope to enable and what it means for todayís commercial processor giants such as Intel, ARM and Imagination Technologies.