Intel Loses Steam Thread

With Intel laying off 5% of its workforce and delaying opening of its AZ 14nm foundry, perhaps it's time to start an Intel Shows Signs of Strain / Execution Gloom Thread.

Even Intel is now succumbing to the massive costs of process refinements and seeing the need to open itself to fabbing for third parties as well as slowing down Tick-Tock with the delay of Broadwell. Meanwhile, the likes of TSMC have now had years of experience as a foundry partner, and Intel is a fledgling with its own design rules that are only familiar to itself.

Atom was a second class citizen at Intel and languished for far too long while Intel focused on pushing its high end far past the knee of the curve when it comes to performance per watt. Atom was a barely acceptable as a netbook part and at the same time, too power hungry to be a great mobile platform. As a result, ARM compatible systems are now dominant, with the added plus of flexibility for custom implementations. Apple's A7 shows very impressive performance and interesting design choices in on die cache for instance that wouldn't be as exclusive or custom had they gone to Intel instead.

Xeon Phi has seen extremely limited acceptance so far compared to the relatively well established GPU architectures. The new Silvermont module will be the basis of future designs but we shall see if this is enough to find acceptance and keep existing GPU architectures at bay.

I also wouldn't be surprised to see a full stop to money losing ventures like Itanium in the near future in an effort to shore up now limited resources.

 

Apologies for the unclear title, I wasn't aware of any existing deal Intel had with Valve or any reason why Valve wouldn't support Intel platforms. It's not meant to be the same as the AMD thread (hence "Losing Steam"), however the new reality seems like Intel is slowing down and executing in a slowing PC environment and a upstart in mobile rather than head and shoulders above anyone else in a PC growing environment. It's not anywhere close to AMD in terms of finances, I agree; earnings last quarter were merely disappointing, not negative.

There are some differences between the environment Intel faces now and what it faced back with its x86 competitors. It doesn't own the instruction set anymore and the modern paradigm seems to be to design your own SoC against open standards and to find your own fab. I'll point to Apple's excellent A6 and A7 as poster children of this, and I doubt the likes Apple would really want to revert to a model where Intel owns a lot more of the intellectual property present in their silicon. The one positive rumor that I've heard in Intel's favor is that the next Nexus will be based on a Silvermont SoC. JITers are making it easier to be hardware agnostic though there probably will still be compatibility headaches. Intel doesn't have monopoly status in mobile and though the market is growing, its margins will be thinner here, and even though it will do respectably here, it's hard to see how it will seize a dominant position here going forward.

If Intel opens its foundries in the future, it transitions to a less profitable mode of filling its capacity since it's not getting the profit at the point of sale anymore, and I expect its 1 node advantage over the rest of the industry won't last forever especially with slowing profit margins and dwindling expenditures on new capacity; the transition to 450 mm wafers will also be expensive. It will gradually be one among many peers rather than head and shoulders above the rest in the foundry business although certainly a force to be reckoned with.

I don't see Intel losing its dominance for traditional PCs and servers, though microservers and APUs are chipping away at its market share and margins in this space. Intel will be solidly profitable based solely on this, but last quarter we saw the effects of this new reality on its bottom line.

 

Intel's process dominance in the past doesn't guarantee its dominance going into the future. Continuing to refine the foundry process requires ever increasing investments which they're starting to postpone. The missing income from sustained growth and additional capacity will likely hurt in the future when very large investments are needed to develop around hard physical limits past 10nm that everyone will face. Intel themselves have said that they see a "clear path" to 10nm, but it'll be challenging even for them past that and their relative advantage now might seem smaller in the face of a physical limit. And though Intel is ahead process wise, its competitors are not as far behind as the name of their nodes seems to indicate. (See: http://www.chip-architect.com/news/2010_09_04_AMDs_Bobcat_versus_Intels_Atom.html) It's looking more like Intel is sitting amongst peers as time goes on.

As for mobile, it won't enjoy monopoly pricing power in mobile or have its shady "rebates" program to sway customers going forward. When it comes to SoCs, other companies like Qualcomm, Apple, and AMD have just as much expertise and excellent existing products based increasingly on licensed 3rd party IP like ARM CPUs and PowerVR GPUs. Intel will certainly be very competitive, I seem to remember interesting technologies like on chip, CMOS modems, but I seriously doubt it'll achieve the heights of the Intel of old since it will have to make do with its sliver of the mobile pie.

 

That's fair and thanks for the heads up to the RWT thread. Intel will be solidly profitable and a leader in process for the foreseeable future, but we're getting well into diminishing returns when it comes to fab investments and keeping the same distance from the field will start to get harder and harder especially without the growth in income Intel was expecting. I think these rumblings mean that it is unlikely that Intel will enjoy the kind of dynasty it had for x86 during the 90's and 00's going forward.

As for monopolies, I have nothing against benevolent and efficiently run monopolies but of course, Intel was anything but that.

 

BTW, are you talking about the thread root entitled: "14nm Fab 42 killed before birth?"

 

I see that the title really is causing some cognitive dissonance!