Intel will start building ARM-based smartphone chips

Intel CEO Brian Krzanich at the Intel Developer Forum today.

Intel has entered into a new licensing agreement with competitor ARM to produce ARM-based chips in Intel factories. The deal, announced today at the Intel Developer Forum, is a strategic move from the Santa Clara, CA company to offer its large-scale custom chip manufacturing facilities, which include 10-nanometer production lines, to third-parties, including those using its rival's technology. Intel's new foundry licensing agreements also include deals to produce chips for LG Electronics, Netronome, and Spreadrum.

"I think it can make a real difference in the industry," said Will Abbey, a general manager in ARM's physical design group, at IDF today. "For me, it simply makes sense." Intel will be tasked with producing ARM's 64-bit cores and other hardware that companies like Apple, Qualcomm, Nvidia, and others base their chip designs off of. ARM-based chips power a majority of the world's mobile phones, with many companies relying on either their own designs or those developed by Qualcomm.

The deal is another reminder that Intel missed the boat on mobile. The company ceded ground to Samsung, Taiwan Semiconductor Manufacturing Co., and other chip makers who now assume the dominant and lucrative chip making role for smartphones that Intel has enjoyed with PCs for decades. Now, with PC sales declining, Intel is eager to make use of its foundry infrastructure to bring in new customers, even if it means developing chips it once hoped to design itself.

Though it seems like an admission of defeat from Intel, the move is not altogether surprising. Rumors began percolating a few years back that Intel was in talks with Apple to manufacture chips for the iPhone and iPad, both of which use processors with ARM-based architecture. For years, Apple has been forced to turn to Samsung and TSMC to produce its Ax chip line. So it makes sense the iPhone maker would be interested in reducing its manufacturing dependency on its central smartphone competitor.

With an ARM deal in place, Intel now has the licensing framework to grab its share of the iPhone chip business, pending of course any concrete deal with Apple. Given the proximity of a new iPhone launch next month and reports that Apple has already inked a deal with TSMC to produce the A11 chip for the 2017 iPhone, Intel may have to wait a while before it gets in on that business.

Comments

I can see Apple switching to Intel for manufacturing.

I’m sure it is already in the works. Intel would have sounded out potential customers before taking the license, and Apple is the biggest and most obvious. It wouldn’t surprise me if Apple had been asking Intel to do this.

One needs a new supplier; the other needs new custom.

But this news has nothing to do with Apple, actually.

Apple designs their own SoC cores, and the news is that Intel will re-enter the ARM chip business that they already had and killed off (StrongARM line) years ago.

Whether Intel will be ready to work as contractual manufacturer for Apple or not is a totally separate question. If they wanted, they could be doing it for years already, but they did not as it is a low margin business. There was no point for Intel to gave up their manufacturing facilities for products that bring so little. And Intel does not need to license anything from ARM to be a contractual manufacturer; the IP is licensed/owned/managed by Apple.

So what we have here is the news that Intel can become competitor to Qualcomm in the market of SoCs for Android phones and many more other devices, not the news on anything Apple-related.

The question on whether Intel will become contractor manufacturer for Apple is no, and this news is another confirmation of that. Intel is going to sell/design ARM SoCs which will bring them way higher margins than being just a contractor manufacturer as TSMC.

Intel has a competitive advantage because their manufacturer process is finer than that of TSMC, Samsung Semiconductor, Global Foundries even at the same nominal 14 nm norms. This means that their ARM SoCs can be be more competitive than whatever Qualcomm or Samsung can get. (Apple is out of the competition since they only need SoCs for themselves.)

Article quote:

offer its large-scale custom chip manufacturing facilities, which include 10-nanometer production lines, to third-parties, including those using its rival’s technology

This is not selling their own ARM chips.

Yes, I stand corrected, thanks; my bad; I just looked at the source again and it refers to specific ARM cores only as example that it is tied to the underlying related IP that Intel licensed from ARM, but not the cores themselves.

The point that the news is not related to Apple that does not use any of that IP (which is tied to ARM’s cores) still stands. Intel did not need to license anything to become Apple’s contractor manufacturer as Apple licenses ISA, and does both high level and low level design in-house.

The news is Intel will manufacture specifically ARM-cores-bades SoCs for LG Electronics, among other ARM-related products.

By the way, one of the reasons why Intel still did not become Apple’s contractor manufacturer is not only because it is purely economically less profitable for Intel than to make chips for themselves, but the question of scale.

Intel opened up for third parties only three years ago and until now their projects were tiny. The signed deal with LG Electronics will be by far the biggest for Intel’s Custom Foundry (as they call it). Yes, this contract offers bigger margin that possible contract with Apple due to the ARM licensing terms that allow Intel to sell those chips to LG Electronics for a good price, but, more importantly, even this biggest deal is more than order of magnitude smaller than anything that could potentially interest Apple (they need a couple of hundred of million SoCs per year).

Thank you for your informative comments.

We don’t know that. There are lots of reasons that Intel might not have been willing or able to supply chips to Apple. Not to mention the fact that Apple never talk about who their supppliers are, and have taken steps in recent models to obscure the exact manufacturer of your components (so they don’t get people returning, say, LG-manufactured displays in favour of a Samsung-manufactured one). There is a special "ARM Foundry License" — perhaps there is IP they need to license just to fab the chips correctly (e.g. securcore).

If Intel started supplying Apple, the first place we’d see it is in revenue guidance. They’d talk vaguely about a large customer. Then Apple would announce the chip, not saying who fabs it, but marketing advanced silicon features which would give it away.

I listed all of the reasons.

If Intel will be able and willing to be Apple’s contractor manufacturer, it is not going to be even next year, 2018-2019 would be the earliest due to issues with scale. The manufacturing equipment purchase plans and accommodation for the future is planned long before actual large scale production could start.

The news has plenty to do with Apple – Apple design the overall SoC but don’t manufacture anything. There’s plenty of scope for them to lean on Intel, especially if they need to start producing differentiated products for other lines, such as the Watch, iPads, ARM notebooks and other potential future products.

Also, until TSMC and Samsung catch up with Intel’s 10nm process, which would be highly advantageous for phones, there’s always the draw.

Why would apple need a new supplier? TSMC seems more than capable.
Though, to me it makes sense as well that they would announce this only after catching a few big fish.

The key advantage will be being able to build in the modem chips to the SOC. It appears Apple are using Intel rather than Qualcomm modems in iPhone 7. Next logical step..,

Intel has the best manufacturing process among all the competition: TSMC, Samsung Semiconductor, GlobalFoundries. But, as I explained above, Intel is not ready to become Apple’s contractor manufacturer due to economic reasons as well as scale. Otherwise Apple would totally love to use Intel.

Even if Intel wanted to do this, I wonder if they’d be able to do it cheaper than TSMC and Samsung.

Intel is going to sell/design ARM SoCs

No they are not. They are going to be a fab for others.

Remember when Intel sold their ARM stuff about a decade back to focus on x86?

I heard that when they bought Infineon and were forced to take an ARM license, that they got a pretty expensive deal (as you would expect – ARM isn’t going to give Intel any discounts).

They probably took a cheaper option this time. Maybe a fab-only license?

And does Adreno rearranged spell anything familiar?

cough Radeon cough…AMD in their infinite wiseness sold it right before the smartphone boom

Those were desperate times for AMD.

They did and they won!

Now they’re coming back to ARM! What’s wrong with that?

This will substantially bring things forward.

Everyone elses "14nm" is what Intel would call 22nm with Finfetts. It’s marketing spin and each fab counts transistor size differently.

(notice the part where TSMCs 20 to 16nm doesn’t even change size?)

As you can see, Intels lead on the world continues. Apart from size also leakage and other features of fabs are still things Intel is ahead on. Their fab advantage didn’t shrink, or at least not as dramatically, just TSMCs/Samsungs/Glofos marketing decided to jump ahead of line.

Intel also uses the newest process on both front end of line and back end of line transistors, while the others use one process back from current on back end of line.

So ARM managed to oust Intel from smartphones despite that huge disadvantage. WITH Intels advanced fabs…Things will be interesting.

Technical lead <> Market Lead

Honestly, the size isn’t all that matters. TSMC’s 16nm setup already outperformed Samsung’s 14nm LPE in terms of yield and power consumption (see the ridiculous Apple A9-chipgate), the smaller die size really only matters in terms of how many chips you can get out of a wafer but at this point all these chips are so small that the cost savings gets incrementally smaller since you can already pack so many chips into each wafer. It’s not like the old days where moving to a new process node yielded a huge decrease in the cost per chip.

Intel might indeed have the superior process though, but they might also not be as good as TSMC at helping to tape out other people’s designs, since they’ve never done contract manufacturing work before. And they also have more of a conflict of interest, since they’re still a chip design company I could see a lot of fabless semiconductor companies being more reluctant to give Intel business than a foundry like GloFo or TSMC where they aren’t potentially also competing against you. Apple uses Samsung but they obviously are trying to move away from that and I’m sure the fact that Samsung designs it’s own SoCs that directly compete with Apple’s is part of that.

I guess we’ll see if intel can convince people that they won’t end up being a competitor someday.

"TSMC’s 16nm setup already outperformed Samsung’s 14nm LPE in terms of yield and power consumption (see the ridiculous Apple A9-chipgate), the smaller die size really only matters in terms of how many chips you can get"

See the image above, Samsung/Glofos 16nm isn’t even smaller than TSMCs 14nm, like my point was, it’s marketing.

As I also said Intel leads on not just die size, but transistor energy leakage and other power characteristics. This is well documented.

Whether they can be as good partners as TSMC remains to be seen, but there would certainly be big advantages to successfully moving to their fab.

View All Comments
Back to top ↑