Intel has 1,000 people working on a chip for iPhone? Of course they do

Late last week Venture Beat reported that “Intel has 1,000 people working on chips for the iPhone.” This prompted a bunch of inbound e-mails and a Twitter conversation. After a few back-and-forths, I realized that I was never going to express what I wanted in 140 characters, and so here we are with a blog post. To put it concisely, of course Intel has 1,000 people working on chips for the iPhone. I would not be surprised if they had many more. The article mentions some interesting tidbits, but there is nothing is really new going on, and I think it is hard to make big calls on Intel’s strategy based on the simple fact that Intel has a project to build LTE modems.

First, let me be clear on something. Everything I am writing here is based on press accounts, publicly available materials and brokerage reports, plus a dose of common industry practice. I have no inside knowledge of what is going on between the two most secretive companies in the country. All of this is outside information.

Intel bought the wireless business out of Infineon in 2011. Prior to that Infineon had a good wireless business. They probably made the best radios (i.e. transceivers) on the market, and had very good modems as well. They had both of those chips in the original iPhone versions. However, the modem business is incredibly competitive. And that wireless business required more resources than Infineon could provide. Intel has plowed a lot of money into the business since the acquisition since then, and Infineon probably could never have afforded it. Intel bought that wireless business and set the team to keep doing what they had already been doing. If they have a 1,000 people working on mobile chips, they probably had the same number last year and the year before. This is Intel investing in a wireless strategy.

Second, competing in the modem or baseband business is punishing. It requires billions of dollars of investment every year just to keep up. In addition to all the demands of a typical semiconductor design process, wireless products for cellphones have to keep up with the mobile standards – namely 3G, 3.5G, 4G (or to be precise GSM/GPRS/EDGE, CDMA, UMTS, HSPA, LTE, etc.) Those standards keep moving forward, and chip suppliers have to be constantly investing in software teams to keep up to date. This process got worse in the transition from 2G to 3G, and even more intense in the transition from 3G to 4G. Today, there are effectively four vendors of ‘merchant’ silicon, that is basebands which are for sale to external buyers, plus two internal efforts at two large handset vendors (Samsung and Huawei). Ten years ago, there were at least a dozen merchant vendors. So the technology has gotten much more complex, the R&D demands have gone way up and consequently the industry has consolidated.

To make matters worse, the customer landscape has altered dramatically. I touched on this subject last week, and called the industry contorted. There is Apple at the high end, and they are essentially the only profitable handset vendor out there. There are a shrinking number of global brands like Samsung, Huawei, Sony and HTC. Most of these are struggling as they are stuck between Apple’s dominance at the top of the pyramid and then a sea of low-cost companies coming out of China.

Intel has said they want to be competitive in mobile. They have been backing that up with literally billions of dollars of subsidies and marketing support for their customers. The trick is who will buy their products. Going after the low-end is incredibly difficult. Mediatek and Spreadtrum have huge teams of engineers supporting the China handset makers, in a ruthlessly price-sensitive market. Intel could go after the global brands like HTC or Sony. While I have not done the math, my guess is that the collective market of all of these is probably not big enough to cover the investment required. That leaves Apple as the biggest, and possibly only, opportunity. As I said in my piece last week, winning Apple is hard, but Intel has to make an effort. This is the problem that all component vendors face – there is really one sizable customer left, and that customer is aware of their position, and negotiates accordingly.

So Intel is making a bold effort. They are the best positioned company out there to win some share of the iPhone modem business, because they have the scale to maintain the investment required. The analyst and investor community generally believe that Intel has won some share of the modems for the iPhone 2016. The VB article points out that nothing is certain yet, and Intel still has to get their product to the finish line. I get the sense that most analysts believe the Intel program has been behind schedule, but seems to be very close to actually getting the chip working. Intel also passed a big hurdle earlier this year when they acquired the CDMA assets from Via of Taiwan. Any modem for the iPhone has to support the CDMA standard, which is used by about 15% – 20% of phones today. Effectively, for years only Qualcomm and Via could do this, and now Intel can (and probably Mediatek, though that is a bit murkier). So Intel is close to having a truly global LTE modem, and Apple is the natural customer for the effort.

I think this is all pretty straightforward, the trouble is making a call on Intel’s strategy from this set of ‘facts’. Intel faces a challenging world. PC sales have peaked, and while they will not go away entirely, the company’s core market is definitely getting smaller. Their investment in mobile is part of their strategy to address this. I do not think that we can draw any further conclusions than this. At least no conclusion justified by the fact that they want to win some Apple business.

Intel is also confounded by the growth of ARM processors. Put simply, this refers to a whole ecosystem of chip companies. These license certain parts of their chips from ARM and use outside fabs, or foundries, for manufacturing. By contrast, Intel is vertically integrated. They do all their own chip and processor design, then manufacture chips in their own fabs. The people at ARM argue that ecosystems always win because the diversity of suppliers allows for a much more nuanced diversification strategy. Fabless companies designing ARM-based chips can cover every niche. And if you want to be precise, the Infineon modem Intel is building is probably ARM-based. I am sympathetic to ARM’s argument, but we also have to recognize that Intel still has the best fabs out there and immense amounts of talent and resources. If you want to be abstract about it, no one can argue that Intel’s vertically integrated model is unworkable while at the same time arguing that Apple’s vertically integrated market is destined to succeed. There is no fundamental reason why vertical integration is somehow inferior, a lot depends on other factors.

In fact, I would argue that Intel has a lot of levers they can pull to win Apple as a wireless customer. They can offer a bundled deal which includes processors for the MacBook, and even cut Apple a deal to serve as their foundry for future versions of Apple’s A-Series of processors. I have no idea if any of these will ever happen, but I want to point out that this is a complex negotiation environment.

And, of course, there is Apple itself to reckon with. The post makes a big deal about the fact that Apple hired a big team of people from Infineon, but that started years ago, and that team has been using Qualcomm modems for a long time. More intriguing is the idea that Apple just wants to license the modem software from Intel and then design their own chip. That rumor has been circling for a long time. And I think it is important to remember that. Apple wants to manage their suppliers for its own ends. They now have two foundry partners to fight over iPhone share. For the past few years Apple has had little choice but to use Qualcomm for modems, so it is only natural for them to want a second source. When (if?) Intel finally gets its LTE modem working, Apple will have that second source. My guess is that Apple really does not want to design its own modems. That requires a lot of labor intensive software work to keep up with those standards mentioned above. At this point, Apple could definitely go out and hire all those people. However, it may be more efficient for them to simply push their two chip vendors to do the work and license the end-product, or just buy the cheapest silicon they can negotiate. Personally, I think Apple’s best supply-chain skill is the way it leverages the R&D resources of its vendors.

In the end, I do not know what will happen here, but the industry landscape is already pretty ugly. I think both Qualcomm and Intel are in a difficult position with regards to Apple. But then again, so is every other chip vendor.

10 responses to “Intel has 1,000 people working on a chip for iPhone? Of course they do

  1. I’m shocked SHOCKED to hear Intel would have a lot of people allocated to trying to do business with the world’s most profitable handset maker.

  2. Pingback: Start up: Safe Harbour’s failure, Google Photos grows, Android Doze, Theranos redux, and more | The Overspill: when there's more that I want to say·

  3. Pingback: Should Intel Split in Two? Probably Not – DIGITS to DOLLARS·

  4. The interesting factor not mentioned is the Apple Watch. Eventually that’s going to want a SoC with integrated modem.

    • Agreed, but all the RF people I talk to warn that it is going to be a long time before we can get a modem small enough to fit in a watch. not just the chip but all the antennas and other components. I imagine a lot of people are working on that, but personally, I think it is still years away.

      • I would have thought so except there are already smart watches with integrated LTE modems. Granted they are much bigger than the Apple Watch but still…

  5. There are a few things missing from this piece. And so many things wrong with almost all internet report on the matter. But sometimes that is to be expected, the Apple communities isn’t exactly the most tech savy bunch. Anandtech used to provide a lot of these information, but now Anand is working in Apple i guess may be we dont see these review anymore.

    The first thing, the iPhone 6s is using Qualcomm MDM9635, NOT 9645 as some has mistaken. I have no idea why many site got this wrong. As Apple has a tradition to use an generation older Qualcomm chip rather then latest. To me this is mostly a yield issue. After all iPhone, as well as some iPad is expected to reach 200M unit per year.

    The Intel 7360 has been in the work for a long time. Unless Intel has been secretly working on a alternate version, the 7360 is still based on TSMC 28nm. While the next generation Qualcomm MDM9645 which is now renamed to X12 LTE modem is based on TSMC 20nm. The MDM9645 is also a little more advance, supporting higher profile with LTE 600Mbps compared to 450Mbps in the Intel 7360.

    This means unless Intel has some magic, it is highly unlikely the Intel 7360 will be as power efficient as the MDM9645 / X12 baseband, or even the current MDM9635 in the iPhone 6s, both on TSMC 20nm.

    Intel 7360 should be cheaper, since 28nm is likely the cheapest node for a long time.
    Intel has a 7460 on the roadmap long time ago but I haven’t seen any leaks and update from them. The 7460 will be Fabbed by Intel on 14nm FinFET. And should offer similar feature set to Qualcomm 9645.

    The most likely scenario, ( If Apple decide to use Intel Baseband ) is Apple switching to Intel 7360 in iPhone 7, being a new iPhone design the BOM tends to be highest, the cost saving from Modem Baseband could help offset that. And in the 7s Apple will be using the Intel 7460 on their FinFET.

    To end with a few note.
    Baseband isn’t just about technical difficult, it is also about patents and royalty. Since Qualcomm gets their fair amount from Apple i wonder if there are any other financial incentive to keep Apple with them.

    There is no reason why Apple cant design their own Integrated Baseband. Mediatek, Samsung and HiSilicon’s LTE modem has one thing in common, they all uses CEVA LTE modem design. I fail to see why Apple can not licenses the CEVA IP portfolio as well. It will still take a lot of work, but perfectly within Apple’s reach. ( But I dont think any of those baseband modem are any where as good as Qualcomm, I do think Intel / Infinoeon will be competitive based on their pass reputation. )

    The interesting side effect ( Or even main purpose ) for Intel baseband on TSMC 28nm and 14nm on their own, is how Apple balance out the Fab capacity play. TSMC has been having 20nm capacity issues, they have literally converted their 20nm to 16nm for Apple. Apple using Intel 14nm for baseband will means TSMC have additional capacity to offer to other client, hence not forcing them out to Samsung.

    • Thanks for the comments. There is a lot in there, and you clearly know a lot about the industry.
      I think you make a particularly valid point that I skipped over in my piece. Technical specs and cost are just part of the equation, IP is also very important. You mention the CEVA DSP stack. That is one part of it, and it is a good question as to whether or not Apple has a DSP license from anyone, or if they, like Qualcomm, have gone out and built their own. But more than just the DSP stack is the broader issue of licensing around all these technologies. I do not want to get into all the details here, but it is an important, sensitive subject for all the players and will likely be an important negotiating leverage in any of these conversations.

  6. Pingback: LessThunk.com « Intel has 1,000 people working on a chip for iPhone? Of course they do – DIGITS to DOLLARS·

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s