Why Intel Keeps Losing Money In Mobile (Op-Ed)

Intel announced that it's going to change its financial reporting structure by combining the operating results of its Mobile and Communications Group and its PC Client Group. From now on, Intel will be reporting its financial results only for the combined Client Computing Group, rather than report separately its mobile and its PC results. This change will be applied starting with the next earnings report on April 14.

According to the company, the new group was created to "address all aspects of the client computing market segment and utilize Intel's intellectual property to offer compelling customer solutions."

That makes some sense, as Intel, much like Microsoft and other tech companies, are now trying to unify their mobile and PC platforms. However, one consequence of this is also that shareholders as well as others may not be able to see as clearly how much money Intel is losing in mobile anymore.

Intel has been losing roughly $1 billion per quarter for the past two years (about $8 billion in total). That's an incredibly large sum of money for any of its competitors, whether we're talking about AMD, Nvidia, Qualcomm or other ARM chip makers. Those sort of losses would be unsustainable for them, but because Intel still makes so much money from its PC and server chip divisions, the company is able to eat that cost.

High Cost Structures

Why is Intel losing so much money in the first place? There are multiple reasons. One is because Intel used to be primarily a PC chip company, with higher cost structures. Higher cost structure basically means "cost bloat." That's not necessarily a bad thing in and of itself. PC chips have evolved a certain way and from a certain technology base (CISC) that has led to an increased base cost for designing such a chip, at least compared to ARM mobile chips.

Intel's x86 chips still contain some CISC baggage compared to RISC chips, which is made evident by the fact that ARM chips on one- or two-generations older process nodes can still be competitive against the latest Atom chips. If ARM chips were actually built on the same Intel process node, they would likely be far ahead of Atom chips.

To be competitive with ARM chips, Atom chips need "something more" such as other instructions, or a newer more expensive process node, and so on. All of that adds to Intel's cost structure for designing a mobile chip.

Intel has also enjoyed significantly larger profit margins than mobile chip makers. While Intel's profits generally measure in tens or even hundreds of dollars per chip unit, ARM chip makers usually measure their profits in single-digit dollar figures or low two-digit numbers at best for every chip unit. When Intel applies those high cost structures and profit margins to its mobile chips, it discovers that it can't be competitive in mobile with its "normal" price points.

This happened before with Intel's XScale division, and it's likely the reason why it was forced to eventually sell it and get out of the mobile market. Even though there are plenty of profitable (ARM-based) mobile chip makers in the industry, Intel couldn't be one. If it made any profit, it likely wasn't sufficient for a company such as Intel to justify the existence of its mobile business.

Heavy Subsidies

Another reason for why Intel is losing so much money is because to get to the price point where it's competitive against high-end ARM chips, the company needs to heavily subsidize its Atom chips.

An Atom chip would normally sell for over $50, but Intel can't compete at that price point (in the mobile market) because that's roughly twice as much as most ARM chip makers charge a similarly-performing high-end chip. Intel wants to be in the mobile market, so at least for now it's willing to eat that subsidy.

However, the ideal price point for Atom in the mobile market would be somewhere close to $80, if not more. Intel has already begun selling Atom-based chips under the brand name of Celeron and Pentium for $82 to $161. Intel doesn't seem to want to sell Atom chips as low as $30 (or less), which leads us to reason number three.

Pricey Partnerships

A third reason why Intel is losing money in mobile would be that the company is investing heavily in other areas of its mobile chip business to get into the market. That includes partnership programs (making it easier for OEMs to use Atom chips in their devices, paying for branding, etc.) or investments into other mobile chip companies in order to get them to use its Atom design in their SoCs.

This third reason is actually a consequence of the first (high cost structures). Instead of trying to sell $30 chips on its own at a loss, the company would rather license the Atom design to makers of inexpensive SoCs, such as Rockchip and Spreadtrum.

Intel seems willing to do this even though the royalty from those chips will at best gain Intel only millions of dollars a year (even ARM itself doesn’t make that much per year, and its IP dominates the mobile market). However, millions of dollars in profit is still better than billions in losses.

Another important thing to note here is that Intel is also outsourcing one of its core competencies, that of using its advanced process technologies to third-party foundries such as TSMC in order to build those inexpensive Atom-based Rockchip SoCs.

Intel wouldn't do this if it could profitably build its own Atom chips for the low-end on its own 14 nm, 22 nm or even 32 nm process nodes. Instead, it seems the better outcome is to have partners such as Rockchip and Spreadtrum build mobile chips on a competitor's 28 nm planar process technology (TSMC).

In its recent press release, Intel said it would post an $800 million turnaround in mobile this year. (That means posting a $200 million loss instead of $1 billion.) However, now that Intel is combining its reporting, it may be more difficult to parse out how mobile is doing compared to PCs.

If the company was losing $1 billion per quarter for the past two years, it must have had some big money-bleeding programs that it's now willing to cut. Those programs could include the subsidies, the partner sponsorships and so on. 

So the question is, with those programs gone, will that hurt Intel's growth in the mobile market, or is the company confident enough that its chips are on an unstoppable growth path regardless of their lack of subsidies?

Does Intel have another strategy up its sleeve that can offset $1 billion in dollar per quarter in mobile investments? Does that include counting the new $82 to $161 Atom-based Celerons and Pentiums, which mainly will be used in low-end PCs and Chromebooks, as mobile chips?

If that's so, that would explain a rapid shift from billion dollar losses to something closer to profitability, as the company could essentially shift the extremely high profit margins from $161 Atom chips to the subsidized $30 Atom chips for smartphones and tablets. It's not clear how exactly Intel is going to make up that $800 million this year, which is why these questions still remain unanswered.

Regardless of whether Intel will actually post better financials in mobile soon or not, it's obvious the company wants to play this game for the long haul. Uniting the mobile and PC divisions could create some synergy, making it easier to transfer technology between the two, but the effects of that probably won't be seen for another few years.

Follow us @tomshardware, on Facebook and on Google+.

Lucian Armasu
Lucian Armasu is a Contributing Writer for Tom's Hardware US. He covers software news and the issues surrounding privacy and security.
  • TechyInAZ
    While I like Intel's capitalism. I think they should stay out of the mobile market. If they keep failing like this and loosing money, they should just stick to the CPU/CPU server market and make better products out of those.
    Reply
  • gsxrme
    While I like Intel's capitalism. I think they should stay out of the mobile market. If they keep failing like this and loosing money, they should just stick to the CPU/CPU server market and make better products out of those.

    Let them fail, It doesn't matter its computation in that market and computation is good. Only if we had real computation in the desktop x86 and server x86 area. AMD hasn't been pushing into at all in years and look what has happened, NOTHING but smaller dies.
    Reply
  • CaedenV
    While I like Intel's capitalism. I think they should stay out of the mobile market. If they keep failing like this and loosing money, they should just stick to the CPU/CPU server market and make better products out of those.
    I disagree, Intel has been bleeding out money in the mobile space the last few years, but it is quickly turning around. Moving from a $1B loss to a $200M loss in a single year is a huge improvement, and they are finding huge ways to reduce chip costs as the technology matures and they get much better yields per chip and they pay off their R&D costs. In another 2 years they should be in a position where it becomes a profitable business.
    While the server market is quite secure, Intel cannot be the big robust company that it is today merely by living off of their server products alone. The desktop and laptop space is their "bread and butter" market, and it is a market that is currently starting to transition out, and will be relegated to a niche market of hardcore gamers and content creation professionals in 10 years. People like you and I may always have our desktops... but in my own extended family I think there are really only 3-4 people who 'need' a traditional desktop or laptop for what they do (and we are talking about 100+ people here). The reason they have desktops is because they are familiar with desktops and are unaware of what phones and tablets are actually capable of today.
    And even for myself it has been interesting to see the transition. I use to have 2 desktops, a laptop, and a server, paired with half a dozen portable devices (from CD players, to phones, to GPS, PDAs etc.). Up until I went back to school last year I had cut all of that down to a desktop, a home server, and a smartphone. With school I added a laptop to the mix, but with 'real' office coming to phones later this year my laptop will become useless again and I will probably sell it for a newer smartphone.
    Not only do people have fewer devices, they hold on to them much longer. I use to upgrade my computer every 2-3 years... now my desktop is coming up on 4 years old and it really just needs a new GPU to get me through the next 2-4 years and I am more than happy with the performance the system gives me. Heck, the home server I am running was my Core2Quad desktop from 8 years ago and it is still running very strong and makes for a rather overkill home server. And I am not alone in this. Most people are dropping the number of devices they keep, and holding onto them much longer... Intel simply cannot remain Intel with this trend.

    But the mobile space is much more important. The mobile space is not just phones and tablets. The mobile space is cars, TVs, receivers, video players, connected sensors, smart toys/gadgets/robotics... it is an extremely large industry that is being birthed right now. Choosing to miss out on that while the traditional desktop and laptop markets phase out over the next 10 years would be extremely short sighted. Anything they can to do break into these markets will more than pay for itself in the not too distant future, and it is worth whatever price needs to be paid to become the dominant player.
    Reply
  • Vlad Rose
    Ever hear the saying, "You can't make an omelet without breaking a couple eggs"? That's what Intel is doing in order to be competitive in the mobile market. The have the financial backing to be able to throw tons of money and research into a project to achieve their end result. Once they've attained that end result, they will finally show profit in that sector.
    This is similar to what Microsoft and Sony have been doing with their console markets. Sell them for a loss to get market share and years down the road make it back in profits.
    Reply
  • de5_Roy
    Intel will make..
    ONE MILLION DOLLARS!! *DUN*DUN*
    from mobile chip sales.

    Atom based desktop boards (the ones with BGA chips) are still rather expensive. and those DDR3 SODIMMs (for atom nuc barebones) aren't getting cheaper either.

    one of the ways Intel can bleed less money if/when ARM inevitably raises license/subscription/royalty fees which has a domino effect over the whole ARM ecosystem.
    Reply
  • edwd2
    at least we're getting dirt cheap baytrail phones and tablets right now. just ordered a zenfone2 for $299 (Atom Z3580 G6430 4GB 32GB 5.5" 1920x1080)
    Reply
  • TNT27
    socket 1150 same generation i7/i5/i3/pentium/celeron all start out as the same basic identical chip, the i7. This is suppose to be not known by public, and is done so to keep costs down for producing chips, its easier and cheaper to produce one chip, then go from there, than create different special chips right from manufacturing.

    They then destroy/blow fuses to limit the cpu's potential, bringing it down to i5/i3/pentiums and so forth.


    They have been doing this forever, even when my father worked with intel, where he was testing/developing the max frequency on p3/p4s hooking it up to car coolant.

    There use to be a huge problem back then of other small companies (mostly foreign companies) buying up huge quantities of cheapo celerons, and then fixing these fuses together to unlock it to what was fastest of its day, the pentiums, this would lead to of course stability issues. Its the same reason the little e, in Intel use to be dropped, to help find false relabeled pentium chips that were previously celeron chips.

    Today, these falsified chips are much harder to produce for those small companies.

    Im not sure on the exact process they do for producing atom chips, but im sure its very similar.
    Reply
  • Brian_R170
    socket 1150 same generation i7/i5/i3/pentium/celeron all start out as the same basic identical chip, the i7. This is suppose to be not known by public, and is done so to keep costs down for producing chips, its easier and cheaper to produce one chip, then go from there, than create different special chips right from manufacturing.

    They then destroy/blow fuses to limit the cpu's potential, bringing it down to i5/i3/pentiums and so forth.


    They have been doing this forever, even when my father worked with intel, where he was testing/developing the max frequency on p3/p4s hooking it up to car coolant.

    There use to be a huge problem back then of other small companies (mostly foreign companies) buying up huge quantities of cheapo celerons, and then fixing these fuses together to unlock it to what was fastest of its day, the pentiums, this would lead to of course stability issues. Its the same reason the little e, in Intel use to be dropped, to help find false relabeled pentium chips that were previously celeron chips.

    Today, these falsified chips are much harder to produce for those small companies.

    Im not sure on the exact process they do for producing atom chips, but im sure its very similar.

    This is not universally true. You can easily tell when two chips are not the same by looking at the die sizes. The socketed Haswell desktop CPUs (LGA1150) have 2 die sizes (130mm^2 and 177mm^2), the dual-core variants are 130mm and the quad-core variants are 177mm). Yes, they use fuses to disable hyper-threading and some of the cache in some chips, but you are wrong about it not being known by the public, pretty much everyone knows this.

    You think the little e in Intel was dropped to prevent faked CPUs? Seriously? That e was dropped in 1968.
    Reply
  • ta152h
    This is an uninformed article, that fails to understand any of the real intricacies.

    Intel is making processors at TSMC because the technology they bought from Infineon has not been brought over to their own fabs, yet. They will be moving all manufacturing to Intel fabs in 2016. So, no, not because of cost, all because they haven't moved all their technology to Intel processes yet.

    Intel had negative revenue in Q4, for the MCG group. Forget the other nonsense about $80 chips, they don't need nearly that much. They generated NEGATIVE money for each sale, so of course lost money.

    Also, the price listed for Intel Celeron and Pentium parts is completely irrelevant. If you look at the price of a Celeron, then look at the price of the same part on a motherboard, you get an idea of how different the list price is from the actual price. The Celeron on the motherboard costs very close to the same. Also keep in mind, a large part of cost can be the GPU, which will vary greatly between a phone and desktop processor.

    So, no, they don't need $80 per chip, or anywhere near it. That's pure fabrication.

    The question is, why would anyone want these chips? Forget the fluff, the reality is, unless Intel essentially gives them away, no one is using them in tablets. AMD's Jaguar/Puma is better in laptops and desktops, so Intel has to price them low to make them attractive. So, there's no area they are better than competing solutions, so Intel has to compete entirely on price. Thus, the losses.
    Reply
  • PaulBags
    If it's not because part of the chip didn't come out right, then they could just make every chip i7 and sell them for less. But I assume it is a yield thing, they fuse damaged parts off in testing; or being fuses they blow themselves in testing. Whatever part of the chip that's left, if it's stable and preforms to a spec, slap that spec model number it and ship it out.

    At least that's what I always assumed when people talk hobbled chips.
    Reply