Intel CEO: Intel open to fabbing for mobile competitors

[NTMBK]

Ashraf did a write up of it here, but the main quotes are:

Q: Do you view you selling your own mobile chips -- which is preferable because, as you've stated before, you get paid for both the design and the manufacturing -- do you see it as an either/or? Bigger foundry effort which would have to, in some way, be driven by mobile because that's where most of the big foundry customers are or Intel selling its own silicon into the mobile space?

BK: No. That market is big enough, our share is low enough, that I think that the shareholders would prefer that we go out and capture as much of the profit and as much as the revenue and margin dollars that we can.

Kind of odd that they would enable their biggest competitors in the CPU market with their fab technology.

 

[coercitiv]

Here's another way to look at it: as long as Intel keeps foundries for themselves, all competing foundries get fed by the rest of the market, and their R&D budget increases accordingly.

If Intel starts fabbing for the competition, the other foundries will have an even tougher time competing, while Intel products will see decreasing R&D costs.

I don't necessarily see this as the better solution, but it does present a different set of perks & risks.

 

[Idontcare]

Ashraf did a write up of it here, but the main quotes are:



Kind of odd that they would enable their biggest competitors in the CPU market with their fab technology.

Basically no different than Samsung's approach to the mobile market. Sure they stand to make more money if they fab and sell their own chips, but they also don't care if they fab and sell someone else's design either.

 

[beginner99]

Here's another way to look at it: as long as Intel keeps foundries for themselves, all competing foundries get fed by the rest of the market, and their R&D budget increases accordingly.

If Intel starts fabbing for the competition, the other foundries will have an even tougher time competing, while Intel products will see decreasing R&D costs.

I don't necessarily see this as the better solution, but it does present a different set of perks & risks.

True. Plus they can always offer a lagging node and still have their own chips proft from the best proccess. Currently I'm sure they could get customers for 22 nm if the price is competitive. AMD GPUs on Intel proccess...that would be cool and ironic.

 

[liahos1]

im sure they are going after apple and not QCOM/MTEK with this type of strategy. The Chandler facility could be a good option.

 

[ShintaiDK]

It would destroy any relevance of TSMC/Samsung for leading nodes. Intel also got much more leading node capacity than those 2 combined.

 

[Rakehellion]

Kind of odd that they would enable their biggest competitors in the CPU market with their fab technology.

If someone's paying you, they're a partner and not a competitor.

 

[Arachnotronic]

Basically no different than Samsung's approach to the mobile market. Sure they stand to make more money if they fab and sell their own chips, but they also don't care if they fab and sell someone else's design either.

I would say that the big difference is that Samsung's home-grown MPU revenue is so feeble and so small a part of Samsung's entire semiconductor ops that it could (does?) conceivably generate more revenue as a fab than as an IDM.

Intel's success hinges on its position as an IDM. That's why Intel needs to be far more careful about whom they let into the fabs than does Samsung, IMO.

 

[EightySix Four]

If someone's paying you, they're a partner and not a competitor.

People can unquestionably be both. Apple & Samsung? Apple & Sony? Apple & Google? Apple & Apple?

 

[witeken]

Ashraf did a write up of it here, but the main quotes are:

Kind of odd that they would enable their biggest competitors in the CPU market with their fab technology.

Their strategy hasn't changed. He gave a more nuanced answer at IM: they do not close the doors for any company, which of course includes smartphone companies. They will not go on a price war, and they can of course reject companies after they've taken a look at Intel's foundry. Intel will of course still be first on their own new process nodes.

Like people often like to say: the transistor is only one factor of the platform.

 

[scannall]

It's a better idea than wrapping their own mobile SOC's in $20.00 bills. At least they'd make money on mobile instead of giving it away. ARM simply has too much momentum in mobile. OEM's don't really have a compelling reason to switch architectures, and a lot of reasons not to.

So, use any surplus fab capacity as a foundry business makes sense.

 

[witeken]

It's a better idea than wrapping their own mobile SOC's in $20.00 bills. At least they'd make money on mobile instead of giving it away.

Are people deliberately trolling or are they simply ignorant about what Intel's contra-revenue really is?

ARM simply has too much momentum in mobile. OEM's don't really have a compelling reason to switch architectures, and a lot of reasons not to.

Switching is as easy as switching from Intel to AMD. You simply go shopping your SoCs somewhere else.

 

[NTMBK]

Are people deliberately trolling or are they simply ignorant about what Intel's contra-revenue really is?

In the last quarter Intel's mobile division literally had negative revenue. They paid out more in rebates and market development funds than they received in payment for their products. However you try to sugar coat it, they were giving away SoCs with a little extra cash on the side.

Why exactly are they still trying to push x86 chips in mobile? They could instead fab ARM chips for QC, Apple, Mediatek or Rockchip instead, and make some honest to god profit from the smartphone market.

 

[Arachnotronic]

Why exactly are they still trying to push x86 chips in mobile? They could instead fab ARM chips for QC, Apple, Mediatek or Rockchip instead, and make some honest to god profit from the smartphone market.

Being a good foundry is something Intel has even less experience at than building mobile SoCs

 

[jdubs03]

I could see Intel fabbing for companies like Apple, but without mobile Intel chips there would be no Windows tablets. So I think they really could do both, they would just need to probably cut Core M's price down a bit so OEM's can hit lower price points with their products.

The contra-revenue program is going to be tapering off with Cherry Trail, so we should see improvement in regards to bottom-line figures, but then again it doesn't seem to be that great of a performer.

 

[scannall]

Why exactly are they still trying to push x86 chips in mobile? They could instead fab ARM chips for QC, Apple, Mediatek or Rockchip instead, and make some honest to god profit from the smartphone market.

Or even go to where the market is, and make their own ARM chip for mobile. They have enough talent to do it. But they'd have to compete with a strong field. Unlike x86 where they just have a weak puppy to kick around.

 

[III-V]

Ashraf did a write up of it here, but the main quotes are:

Kind of odd that they would enable their biggest competitors in the CPU market with their fab technology.

Is this really news? Intel's already gone on record making similar statements.

Regardless, if companies did move over to Intel's fabs, I can't see how it'd be anything but a win for consumers and the world as a whole. Intel's got the best tech, potentially at the best price (or at least, a good price). Imagine AMD moving over -- their process disadvantage would evaporate, and while they'd still trail Intel, they'd be getting better margins and better performance. Obviously it'd be too expensive for AMD to do so, but anyone else would stand to benefit too. Nvidia could actually achieve leadership with their Tegra line. Qualcomm could get back in the game.

Perhaps most excitingly, AMD and Nvidia could fab their high performance GPUs on a process that actually has clear performance incentives, unlike the murky TSMC 20nm.

Perhaps some of the few IDMs left would stand to benefit as well? I am wondering if companies like Texas Instruments, Cypress or Infineon would benefit from dropping their fabs and moving over to Intel (I can't imagine their engineers being happy, though). I think the additional volume would give a much needed boost in demand for 450mm wafer development.

That being said, it's a steep, slippery slope. Intel already has leadership in the area, and the only thing that's kept them from gobbling up the market has been the fact that they just didn't participate in it. A windfall for Intel would be a big blow to TSMC, GloFo, Samsung, and UMC -- and there we move from "big potential for cost savings" for consumers to "now Intel's charging out the nose, just because they can." I can't help but think the benefits would outweigh the potential negatives, though... perhaps someone could sober up my expectations.

Being a good foundry is something Intel has even less experience at than building mobile SoCs

I'm sure they're getting very good feedback from Altera and Panasonic, though.

 

[Fjodor2001]

It would destroy any relevance of TSMC/Samsung for leading nodes. Intel also got much more leading node capacity than those 2 combined.

Yeah, but the thing is Intel is realizing that they will soon no longer have much lead left to TSMC/Samsung/GF, so they might as well sell their spare capacity as a foundry. Remember, Intel's process tech lead is shrinking fast these days.

 

[ShintaiDK]

Plenty of lead. And thats not even including the other technologies.

 

[TurtleCrusher]

I don't see Intel allowing AMD to use their fabs for APUs. Take a 22nm 7850k, keep the GPU clocks as is and the frequency limit on the cores would seriously eat away so much of the mainstream market.

That is if AMD would spend more on marketing than shipping calendars and posters to LAN parties.

 

[seitur]

Ashraf did a write up of it here, but the main quotes are:



Kind of odd that they would enable their biggest competitors in the CPU market with their fab technology.

Intel competitors would indirectly fund Intel chip selling business.

Seems like win to me.

Same as Apple fabbing their chips and other stuff at Samsung means that Samsung get more % of overall smartphone cake and help Samsung mobile business.


More companies fabbing at Intel means TSMC/Glofo getting less revenue which again help Intel foundry business.

Intel getting their mobile competitor orders for their fab is nothing but win for Intel.


In the long-run it may not be good for us as consumers though. Especially if Intel success would mean demise of other foundries.

 

[jdubs03]

As long as Intel can retain their wafer prices, and if the competitor's order volumes can overcome an reduction in Intel's own Atom orders (if that is the case, doesn't have to be). So yeah it could be financially beneficial for them to let their competitors fab on their leading edge process.

 

[Kuiva maa]

Switching is as easy as switching from Intel to AMD. You simply go shopping your SoCs somewhere else.

Not at all.Android on x86 is still incompatible with some of the most popular play store apps out there.

 

[mavere]

Not at all.Android on x86 is still incompatible with some of the most popular play store apps out there.

It'd help so much if there were a way to tell whether an app is Dalvik/ART-only rather than having everything be suspect.

 

[III-V]

Yeah, but the thing is Intel is realizing that they will soon no longer have much lead left to TSMC/Samsung/GF, so they might as well sell their spare capacity as a foundry. Remember, Intel's process tech lead is shrinking fast these days.

You're just about the only seemingly-neutral party who believes this, you know.