Modest server growth stands out in a flat Q4 2015 for Intel

Intel's fourth-quarter 2015 financial results are in. The company reported revenue of $14.9 billion for the quarter, up 1% compared to Q4 2014, and operating income of $4.3 billion, down 3% from the same time last year. Gross margin shrank 1.1% to 64.3%.

  Q4 2015 Q4 2014 Change
Revenue $14.9 billion $14.7 billion up 1%
Operating income $4.3 billion $4.5 billion down 3%
Net income $3.6 billion $3.7 billion down 1%
Gross margin 64.3% 65.4% down 1.1%

The company's Client Computing Group took in $8.8 billion, a 1% decline compared to this time last year. According to the company's CFO commentary, desktop platform volumes (which account for both processors and chipsets) fell 9% year-on-year, but desktop average selling prices rose 9%. Notebook platform volumes experienced a 10% year-on-year decline, but average selling prices increased 6% on that basis. The company says its tablet volumes fell a precipitous 33% compared to a year ago, but average selling prices were up "significantly."

That news aside, Intel says its shipments of 14-nm processors made up more than 50% of its Client Computing Group volume, and it says its Skylake chips have received an increasingly enthusiastic response from the market.

The Data Center Group raked in $4.3 billion, up 4% compared to last quarter and 5% year-on-year. It appears the strong growth the company experienced in this sector in recent quarters may be coming to an end, though. In Q3 2015, the company posted 8% sequential and 12% year-on-year growth for its server division, while Q2 produced 10% year-on-year growth.

The Internet of Things Group had revenue of $625 million, a 6% year-on-year increase, while the $543 million of revenue from the company's software and services segments was down 3% compared to Q4 2015.

For the first quarter of 2016, Intel expects revenue of $14 billion, plus or minus $500 million, while gross margin is expected to fall to around 58%. Those projections include numbers from FPGA maker Altera, of which Intel concluded its acquisition in late December.

Comments closed
    • DavidC1
    • 4 years ago

    They screwed up with the decision to go “all mobile” at the 14nm generation. Up until 22nm, the foundry processes were more dense. As in, 28nm TSMC would be only 30% larger everything equalized compared to Intel 22nm. Same with TSMC 40nm and Intel 32nm and so on.

    They decided to focus extra on density with 14nm. Bad idea. Remember when Intel/Micron venture for NAND scaled down in size very quickly? Well it screeched to a halt at some point, and others took the lead. There must be a physical limitation for everyone and everything that makes it really really hard. NAND hit it earlier that’s all. Same with memory.

    I think Intel lost the sight of why they had fabrication lead because mobile came from nowhere and they panicked. The number 1 priority for a process is apparently yields and being able to mass produce them. Remember how AMD went copper interconnect first, and also SOI when Intel went conservatively and went later or did not use it at all?* There’s zero point in having high performance process when you can’t produce/yield them. You don’t go fancy because unexpected results make bad things happen.

    14nm should have been a conservative process. They wouldn’t have had mobile anyway, and it would have meant their bread and butter PC chips would have ended up better. Broadwell was 6 months late because of 14nm. Imagine having a 5775C widely available at June 2014. I bet you it would have ended up to a higher performing version than with the 5775C they have now, which is vaporware too. Yield = performance at certain points

    S|A was spot on point with articles like these highlighting Broadwell failures, which are really 14nm related: [url<]http://semiaccurate.com/2015/01/15/analysis-intels-broadwell-financial-struggles/[/url<] *I've also found out that copper interconnect and SOI didn't really make better performing transistors. Of course Intel was masking process advantage with blind sighted efforts like Netburst, but the process was exemplary. AMD was also much late with their process. You don't chew more than you can swallow. You end up not being able to do anything.

      • NoOne ButMe
      • 4 years ago

      SOI can lead to transistor performing better. Or to be more specific allow do higher switching speeds at the same voltage, or lower voltage at the same speed and various combos of these.

      IBM didn’t push SOI for nothing. If you want to argue with the technical prowess of the process engineers IBM had, and some they still have, feel free. The only big downside of SOI I know of is it adds costs. Some people who are crazy about it claim it lower net costs due to increased yields. Doesn’t make sense to me, but could be true for some deigns.

      • Stonebender
      • 4 years ago

      ” had fabrication lead”

      Intel still has a sizeable lead, don’t let marketing tell you otherwise. I don’t expect that to change with 10nm. The multiple patterning that is required at that node is going to give everyone fits.

        • NoOne ButMe
        • 4 years ago

        Who cares if it’s a bit smaller, a bit faster and a bit lower power if it costs 2x more?

        Atom x3 is made at TSMC right now for example (ARK shows 28nm). And Intel say their 14nm is cheaper per transistor than all their process before.
        [url<]http://ark.intel.com/m/products/87460/Intel-Atom-x3-C3445-Processor-1M-Cache-up-to-1_40-GHz#@product/specifications[/url<]

    • TruthSerum
    • 4 years ago

    Modest microcode growth..

    • NoOne ButMe
    • 4 years ago

    So about 700-800 million less income before they cut capital spending on fabs. Intel has had trouble competing with ARM in mobile 2 nodes up. Their fab lead will be gone soon for most metrics. Better hope AMD’s year delay of Seattle defines how the ARM server players will operate forever?

    [url<]http://www.barchart.com/headlines/story/7301944/explaining-intel-corporation-s-capital-expenditure-cuts[/url<] Intel original 2015 plan: 10.5b dollars, cut to 10B, cut to 8.7B. Later I believe cut another to 7.3B [url<]http://www.electronicsweekly.com/news/business/intel-cuts-capex-again-2015-10/[/url<] (10.5-7.3)/4 = 800 million dollars a quarter in added profit. This profit directly slows down their fab investment because as node shrinks equipment gets more expensive. Hence TSMC and Samsung, both spent more, caught up some. Should never have tried to enter mobile. Burns billions of dollars that could have gone towards keeping fab lead, or taking on less debt. But, Intel had to try. If only they had let Steve Jobs bully them into making an x86 CPU that worked for the iPhone. They'll make plenty of money, and I think survive. Just will eventually sell off or shutter most of their fabs. And downsize.

    • TruthSerum
    • 4 years ago

    Hey Chuckula.

    You and SSK can stop trolling people now who point out flaws with Intel/M$ products.

    Thanks,

    The users of TR forums.

      • TruthSerum
      • 4 years ago

      Predictable instant -3 from the fanboy. I guess pride = prejudice, in some circles…

        • Klimax
        • 4 years ago

        Don’t speak for us.

        Thanks, users of TechReport.

        (:D)

      • ronch
      • 4 years ago

      You mean the users of the TR forums collectively use the TruthSerum™ moniker?

    • ronch
    • 4 years ago

    Although Intel is seen as the 800-pound gorilla that’s practically indestructible, remember that Intel has a lot of fabs to keep humming. If global demand tapers off all those expensive fabs will sit around idling, which will kill the company. It’s like Kodak and Agfa, or some other manufacturer of items that were once selling like pancakes then demand suddenly disappeared, killing the company in a flash.

    This is Intel’s biggest and scariest nightmare. They’re ahead but they need to stay ahead or else they’ll die a quick death. This is why they were willing to practically give smartphone SOCs for free to get their foot at the door; they’re THAT desperate to get into whatever sells to keep those fabs running.

    Servers and PCs are going nowhere fast but they’re quickly becoming like washing machines and refrigerators and TVs that people keep for 5-10 years. Intel’s growth and success has largely been because people have kept upgrading their CPUs since the original IBM PC came out, but that trend has been slowing down steadily over the past decade.

    Edit – thinking back, in a sense perhaps AMD made the right move to go Asset Light™. Not that they had a lot of fabs to keep running, of course, but it was even worse for them because they simply didn’t have the economies of scale and silicon revenue to make even that one fab as cost effective as Intel’s fabs. AMD’s dilemma was different from Intel’s but the bottom line is that running your own fabs isn’t as glamorous as it once was.

      • Krogoth
      • 4 years ago

      QFT

        • ronch
        • 4 years ago

        [b<]KROGOTH IS IMPRESSED!!!!!!![/b<] WOOHOOOO!!!! I DID IT!!!!! TAKE THAT, PEOPLE!!!!!!!

      • guardianl
      • 4 years ago

      Agreed. This nightmare has already started for them with the mothballing of the 14nm fab in arizona. Now all of intels good 14 nm products are in short supply (Xeon D, skylake). We also have the 10 nm delay…

        • smilingcrow
        • 4 years ago

        Even with static sales volumes they would need less fab space as the chips get smaller with each shrink. That would naturally lead to the closing of a fab at some point.

          • NeelyCam
          • 4 years ago

          This is what I’ve been wondering for a while. Chips shrink, and you can fit maybe twice as many on a wafer. Sure – it might reduce the cost per chip, but what if you can’t double the demand to sell them all…?

            • smilingcrow
            • 4 years ago

            The question is; how much are the fixed costs for developing a new process node versus the costs of rolling it out across multiple fabs and the other costs per wafer/chip?
            If the initial development cost is relatively small versus the other costs then they can afford to reduce their number of fabs at a particular node without impacting the gross cost per chip/wafer too much.
            Otherwise their margins might take a big hit.

            As well as that , if Zen is successful and powerful leading to inexpensive consumer octo-cores that puts pressure on Intel to make bigger chips to compete and also possibly downward pressure on pricing. So that is a possible double whammy for Intel.

            • NeelyCam
            • 4 years ago

            [quote<]As well as that , if Zen is successful and powerful leading to inexpensive consumer octo-cores that puts pressure on Intel to make bigger chips to compete and also possibly downward pressure on pricing. [/quote<] Don't automatically assume that AMD stuff is going to be inexpensive. The last time AMD chip performed better than Intel's, the price was high. As Intel has most of the PC/Server volume, and AMD desperately needs the money, my guess is that Intel will set the overall market prices, and AMD will adjust Zen pricing based on its performance.

            • chuckula
            • 4 years ago

            If we somehow learn Zen’s pricing before we see the performance metrics here’s all you need to know:
            1. Massively more expensive than what you expect? AMD actually did it, or Lisa Su has barred the doors and is handing out mandatory Kool Aid.

            2. More expensive than what you would expect? AMD might have done it, but they also overpriced Bulldozer on launch too.

            3. Cheaper than what you would expect? I can already predict the spinning and rationalization since I saw the same thing happen after Bulldozer launched.

            • NoOne ButMe
            • 4 years ago

            We won’t know the real Zen pricing. Only the list price for server parts, aka BS, and the consumer prices.
            I expect the pricing structure to look really weird. I think AMD won’t have any parts below Quad Core until the Zen APU hits market. As I Believe AMD has 1 Zen layout.

            • smilingcrow
            • 4 years ago

            Me: “and also [b<]possibly[/b<] downward pressure on pricing." NeelyCam "Don't automatically assume that AMD stuff is going to be inexpensive." A full fat octo-core CPU for a mainstream desktop platform would be a first and if the performance is good enough it certainly gives AMD a chance to price it high. If it competes well against 6 core Intel enthusiast class CPUs but is supported by much cheaper mainstream motherboards they have a lot of room with pricing. Price the CPU high knowing that motherboards are cheaper is one possibility By then Broadwell E will be out so Intel's response might be to adjust pricing v Zen. Interesting times.

            • NoOne ButMe
            • 4 years ago

            You can currently buy a 6C for Intel systems for less than a 6700k. Why no one is doing so and OC confuses me.

            • smilingcrow
            • 4 years ago

            People definitely are doing just that but I suspect mainly only the people that really need the extra cores and those with a large disposable income for IT.
            It’s only recently though that DDR4 has dropped a lot in price putting it close to DDR3 as before that and Skylake was released if you compared say Haswell Refresh with Haswell E 6 core the overall difference in cost at the platform level was very large due to the mobo and RAM pricing.

        • w76
        • 4 years ago

        They mothballed a leading-edge fab? Hmm, you’re right, that’s bigger news than it sounds.

          • Stonebender
          • 4 years ago

          It’s not as big as news as it sounds. They built the fab, but it was never used; tools were never installed.

      • Stonebender
      • 4 years ago

      Kodak went the way of the Dodo because they were replaced by a new technology and couldn’t adapt. Intel does not have this problem. It’s funny how you mention the slow decline of the PC over the last decade, especially since over that same decade Intel has made more and more money.

        • chuckula
        • 4 years ago

        Intel is positioned to weather a storm of lowered consumer demand for silicon better than most other companies I can think of.

        There are MASSIVE economic upheavals happening right now as you can see by the stock market. China is flat-out in a recession even if their government has firing squads out to silence anybody who will admit it. The U.S. Dollar is gaining strength rapidly, which hurts exports, but even given all of that, Intel is still doing quite well.

          • TruthSerum
          • 4 years ago

          You’re officially a fanboy.

            • chuckula
            • 4 years ago

            Better a fanboy than a blithering idiot.

            RING ZERO!

            • TruthSerum
            • 4 years ago

            You might be both, celebrating errata like it’s a good thing breathlessly..

            • Klimax
            • 4 years ago

            Two personal attacks and nothing more? Definitely not good for “your side” if that’s all you have to post here…

        • NTMBK
        • 4 years ago

        Over the last decade Intel has been hoovering up more and more of the server market. PC doesn’t have an awful lot to do with it.

        If they lost the PC volume, it would get trickier to make the economics of owning their own fabs work. Take a look at IBM to see how that goes.

        • TruthSerum
        • 4 years ago

        ” especially since over that same decade Intel has made more and more money. ”

        That doesn’t mean their product has improved. That means they charged more.

          • NeelyCam
          • 4 years ago

          If the product doesn’t improve, people don’t upgrade. Products need to improve to keep the [url=https://forums.geforce.com/default/topic/451468/nvidia-ceo-we-don-t-make-money-cannon-like-intel/<]money cannon[/url<] booming.

          • Stonebender
          • 4 years ago

          Hasn’t improved? So Skylake isn’t any better than a Core 2 Duo? Aside from that nonsense, they haven’t been charging more (at least on the consumer side). In fact, processors in general have recently been more affordable than they’ve ever been.

      • blastdoor
      • 4 years ago

      I think this is generally right, but I’d quibble with some details.

      Having some idle fabs does not represent an existential threat to Intel, though it would be extremely bad for the stock price. Intel wouldn’t go poof like Kodak did. Unlike Kodak, Intel still makes a product people want — they just don’t want as much of it as they used to. A better analogy might be General Motors.

      Among plausibly realistic scenarios, I think the worst case for Intel is that they lose the ability to invest more than the foundries, which would eventually result in Intel losing its edge in fab tech. They would just become another foundry. That would be bad for margins and bad for the stock price, but it wouldn’t kill the company.

      We may one day conclude that turning down the iPhone contract was a bigger mistake than Netburst or Itanium. Imagine if Intel had the iPhone business right now. This would be a very different conversation, and a very different earnings report.

      • w76
      • 4 years ago

      The numbers don’t support the thesis. For them to be in danger of insolvency, there’d need to be a scenario where they couldn’t pay their bills. They’ve got over 21 billion cash, 20 billion in debt. Thus, if it “hit the fan,” they could pay off their debt at maturity (or buy it back or call it if possible) and live off cash flow, with everything having been paid for. They could also scale back R&D and capital improvements, because they’re so far ahead of everyone else as it is and those same competitors would also feel the heat. That’d free up more cash flow.

      If they had a lot of debt, the picture changes, and sure they could end up with ‘stranded assets’, but that’s a problem in the oil industry where they have massive loans against those assets. Intel has no such problems. They could fire-sale property and not be significantly hurt.

      Frankly, most western sovereign governments wish they had such solid financials.

    • kuttan
    • 4 years ago

    I think Intel already maxed out with profits in server segments thanks to no meaningful competition.

    • guardianl
    • 4 years ago

    Intel Skylake-K 4C – 122 mm2 @ 14 nm in 2015
    Intel Ivy Bridge 4C – 160 mm2 @ 22 nm in 2013
    Intel Sandy Bridge 4C – 216 mm2 @ 32 nm in 2011
    Intel Lynnfield 4C – 296 mm2 @ 45nm in 2009

    Notice a trend? Yet Intel’s gross margin has been [url=https://ycharts.com/companies/INTC/gross_profit_margin<]50-60%[/url<] basically the last 5 years. They've been growing their absolute profits on the backs of users with ever smaller dies instead of increased sales volume or real cost savings with new nodes. I'm interested to see what they are going to do now with much slower node development and greatly increased costs plus a shrinking PC market. Unless AMD pulls off a miracle I bet $$$ we see a phone SoC die larger than Intel's quad cores within the next few years...

      • DavidC1
      • 4 years ago

      There’s a limit on how small they can get. If its too small it’ll be like the memory companies that have ever small dies but no volume to sell them. Yea, they are cheap, but in low volume!

      Unfortunately I don’t think this is easily fixable. Cannonlake going up to 8 cores will help, but then what?

      • smilingcrow
      • 4 years ago

      This has been Intel’s recent response:

      “… desktop average selling prices rose 9%. Notebook platform volumes experienced a 10% year-on-year decline, but average selling prices increased 6% on that basis..”

        • NoOne ButMe
        • 4 years ago

        If Zen is at least as competitive as Phenom was I bet that will instantly drop 10% or more in pricing.

      • DrDominodog51
      • 4 years ago

      However, the cost to make the next node has increased significantly overtime. Not enough to compensate for increase in price of mm^2 of die.

      • Stonebender
      • 4 years ago

      I think 10nm is going to be the slowest node development in history. I’m willing to bet that 7nm will come about much quicker, assuming that ASML gets EUVs power issues sorted out.

        • NoOne ButMe
        • 4 years ago

        Maybe from Intel, or hopefully Samsung?

        [url<]http://www.realworldtech.com/forum/?threadid=156613&curpostid=156613[/url<]

          • Stonebender
          • 4 years ago

          Considering how much money Intel has given ASML, I expect they will get EUV up and running first.

            • NoOne ButMe
            • 4 years ago

            But does it benefit Intel to do so at the cost the first mover will deal with? Part of the reason Intel CapEx stood still around 10-11 billion dollars/year was due to stopping absorbing the cost of being the leader to each node. Instead of bearing the full early adopter cost they’ll share it with TSMC, Samsung and possibly a tiny bit with Global Foundries

    • DrDominodog51
    • 4 years ago

    I’m willing to bet the server growth is from Xeon D.

    • chuckula
    • 4 years ago

    Meh. Their profits were only about 2.25 Powerballs.

      • 223 Fan
      • 4 years ago

      What’s that in Apple profits? .33?

      • 223 Fan
      • 4 years ago

      What’s that in AMD profits? -Integer.MAX?

        • chuckula
        • 4 years ago

        AMD would like all profit/loss statements to be unsigned ints.

          • 223 Fan
          • 4 years ago

          Ha ha. Fooled you. In order for the multiplier to be that high AMD would have to be near break even!

          Typical Intel report: “We made a ton of money”. Typical AMD report: “Losses narrow and we expect to return to profitability in the Year of Desktop Linux”.

            • UberGerbil
            • 4 years ago

            The Year of Desktop Linux — does that come before or after Half-Life 3?

            • Neutronbeam
            • 4 years ago

            Desktop Linux is a thing! Not in THIS reality, but yeah, almost definitely somewhere in the multiverse–where Elvis is using his Amiga Linux desktop to ghostwrite plays for Shakespeare. I am almost completely certain that might be happening.

          • derFunkenstein
          • 4 years ago

          they’d probably prefer to submit documents unsigned, too.

        • curtisb
        • 4 years ago

        Cannot divide by zero.

        • ronch
        • 4 years ago

        Well, I’m sure Bulldozer’s 2 X 128-bit FPUs can do it in 400 clock cycles.

    • anotherengineer
    • 4 years ago

    3,600 million in 13 weeks, so 276.923077 million/week, or 39.560 million/day, or 1.648 million/hr That’s a lot of cash.

    1.648 million/hr, that’s a lot of sales per hour!! That’s about 5495 $300 CPU’s an hour!! Wow that’s a lot of silicon.

    I wonder how many tons of it Intel uses per year??

      • ronch
      • 4 years ago

      ($_$;)

Pin It on Pinterest

Share This