AMD just misses breaking even in the fourth quarter of 2016

AMD's fourth-quarter 2016 financials are in. The company reported $1.1 billion in revenue, up 15% year-over-year, and it held operating losses to just $3 million, down substantially from the $49 million it bled away this time last year. For all of 2016, the company reported $4.27 billion in revenue, up 7% over its 2015 take.

  Q4 2016 Q4 2015 Change
Revenue $1.11 billion $958 million up 15%
Operating income -$3 million -$49 million
Net income -$51 million -$102 million
Gross margin 32% 30% up 2%
Earnings per share -$0.01 -$0.10

The Computing and Graphics segment enjoyed a 28% year-on-year boost in revenue to $600 million. AMD says this increase is primarily thanks to increased GPU sales. Even with that boost, the division reported a $21 million operating loss, though that figure is down from the $99 million loss the company recorded in the fourth quarter of 2015. The company also reported higher GPU average selling prices year-over-year.

Meanwhile, the Enterprise, Embedded, and Semi-Custom segment took in $506 million, a 4% year-over-year increase that's attributed to higher embedded and semi-custom revenue. The division reported $47 million in operating profit, down slightly from the $59 million it generated in Q4 2015. The All Other bucket ended up with a $29 million hole in its bottom, up from a $9 million loss in the fourth quarter of 2015, an increase AMD attributes to "higher stock-based compensation charges."

For the first quarter of this year, AMD projects its revenue will decline 11% sequentially, plus or minus 3%. Even with that wrinkle, the company notes that meeting that target would be an 18% year-on-year increase.

With AMD's Ryzen CPUs and Vega GPUs both primed to release in the first half of this year, we're taking bets on whether 2017 will finally mark a return to profitability for the long-suffering chipmaker. Hopefully those critical products live up to the hype.

Comments closed
    • stefem
    • 3 years ago

    AMD looks on track to get out of the hole were they put themself.
    [quote<]The division reported $47 million in operating profit, down slightly from the $59 million it generated in Q4 2015.[/quote<] Jeff, I would not call it "down slightly", that's $12 million less which is 20% decrease in operating profit for the division.

      • ronch
      • 3 years ago

      Yeah. I’m sure $12M is more than enough to set anyone for life. Just don’t throw the money into the fire and you’ll be OK.

    • albundy
    • 3 years ago

    so you are saying ryzen is being pushed back? they said before march, now its first half of this year? good luck to them.

    • AMDisDEC
    • 3 years ago

    Ms. Lisa Su is the best CEO that AMD has had since the times of Jerry Sanders.
    After Sanders retired AMD has had one poor CEO after the other. That last CEO, Rory Reed was the worst CEO you could ever hope to burn money with.

    LOL, Reed reminds me of Bush Jr., the way they both almost destroyed their charges with complete incompetence.

    Ms. Su is focused and at the rate she is going, will have AMD squarely back in the black by 2018.

      • ptsant
      • 3 years ago

      Reed was brought in specifically to do the painful restructuring, laying the ground for Lisa Su. I do not think he was incompetent, he just had an awful task.

        • Anonymous Coward
        • 3 years ago

        I once worked a place that hired a almost-top manager who specialized in wandering the land and wielding the knife. He was actually good at it, I think. Identify waste, slice.

        • ronch
        • 3 years ago

        I think the worst CEO was Hector. He wasn’t a bad CEO per se; he just came in at a time when AMD just had a successful run with the K7 and he just didn’t have the chops to take AMD to the next level after the great Jerry era. At that time K8 was pretty much done so K8’s success wasn’t because of him: he just played the hand Jerry dealt him. We all know what happened after the K8. K10 was just an enhanced K8 and Bulldozer, well. Remember, both K10 and BD projects were started during Hector’s watch and both were OK at best.

        Furthermore, there were rumors that another ‘original’ K8 project was canned but nonetheless the K8 that got released did well partly because Intel shot themselves in both feet with the hot Prescott design and their insistence to define the 64-bit post-x86 PC with Itanium. While K8 was successful, there were again whispers that a K9 project also got canned and Randy Allen and Co. had to quickly hack something to sell, taking the K8 and trying to prop it up for another round and calling it K8L or K10 (K8L die shots look too much like Barcelona cores). Yet again after K10, there were rumors that the original Bulldozer project kinda lost its way and what got released was a toned down, half-hearted, compromised design. All these pointing to difficult times since K7 came out, but at least K8 did well and that wasn’t because of Hector. Jerry got lucky, as usual.

    • rudimentary_lathe
    • 3 years ago

    I would guess that Computing and Graphics would have been very profitable if it weren’t for the antiquated FX line weighing down Polaris. With Ryzen dropping soon and Vega maybe a couple months after, the Computing and Graphics segment should be very profitable from Q2 onwards. It wouldn’t suprise me to see C&G clearing $1B of revenue a quarter by itself come Q3. And if Ryzen and Vega are actually competitive, then considerably more. These are massive markets and AMD has such low comparables yoy.

    • Pancake
    • 3 years ago

    “Annual income twenty pounds, annual expenditure nineteen [pounds] nineteen [shillings] and six [pence], result happiness. Annual income twenty pounds, annual expenditure twenty pounds ought and six, result misery.”

    Charles Dickens

    • tipoo
    • 3 years ago

    “Earnings per share -$0.01”

    And don’t spend it all in one shop!

    • Chrispy_
    • 3 years ago

    I’m surprised they got that close.

    The RX line is selling well and they’re making chips for the new Sony and Microsoft console refreshes, but their CPU portfolio must be really stinging right now. Who in their right mind would buy an AMD processor with Zen imminent? Not only that but their APUs were shown some serious smackdown by Skylake and Kaby has some even more appealing budget options than skylake (I’m looking at you, unlocked i3 and Pentium with hyperthreading!)

      • travbrad
      • 3 years ago

      They do have some good GPUs in the price segments most people care about so maybe that made up for bad CPU sales.

    • ronch
    • 3 years ago

    I’m thinking of buying an FX CPU before it makes its swan song. I won’t open it, it’ll just serve as a memorabilia. I think the Bulldozer core is one of the most interesting CPU designs ever, despite its idiosyncrasies and lack of performance and efficiency and almost killing AMD. See, I’m kinda stuck in performance requirement limbo that if my AM3+ board quits when AM3+ boards are still available, there’s a chance I’d just replace it and not get something newer. It’s cheaper too. And if that new AM3+ mobo died too, I’d have a spare CPU. Heh. Maybe I get to use my 8350 for 10 years.

      • MOSFET
      • 3 years ago

      I’m all for sentimentality. I have an Athlon XP in a motherboard with RAM and video card (not exactly a GPU) ready to roll, whenever needed. Bulldozer is not that interesting. To each his or her own, of course, but I’d rather hang a Phenom II X6 on the wall than an FX-anything. (I’m in a house with two running FX-8320s and an Opteron 6328, garaged and homelabbed, so it’s not like I don’t have extensive experience with AMD since K5. 🙂

        • ronch
        • 3 years ago

        Well, any FX CPU would still be a lot faster than any Athlon XP. It’s actually much more efficient too.

      • DancinJack
      • 3 years ago

      ok…

      • Airmantharp
      • 3 years ago

      Eh, it was like Netburst, but worse…

      Not much nostalgia there for me :D.

        • chuckula
        • 3 years ago

        It was worse than Netburst because when Intel made Netburst they weren’t copying somebody else’s bad ideas. Netburst was Intel’s own homemade bad idea.

          • the
          • 3 years ago

          Then what was the bad idea and who was AMD copying?

            • tipoo
            • 3 years ago

            Long pipelines to chase high clocks. The thermal wall is real.

            • gamerk2
            • 3 years ago

            Intel didn’t know thermals would become such a problem north of 3 GHz. AMD did. That makes BD look a lot worse in hindsight.

            • the
            • 3 years ago

            AMD knew that Bulldozer wouldn’t scale down to ultramobile ahead of time and created Bobcat for that market.

            AMD knew it would be hot, yes, but not [i<]that[/i<] hot.

          • ptsant
          • 3 years ago

          AMD went for multi-threaded performance, which the chip did deliver, and used high-frequency to get single-threaded performance. They lost that bet. Software remains highly dependent on single-threaded performance to this day.

          The design would have scaled and would have been competitive if all that mattered was multi-threaded scaling. That has not been the case and, most importantly, that segment has been absorbed by GPUs. If you have embarassingly parallel algorithms, you make them run on a GPU, not on a CPU.

          I have often wondered what would have been the performance of 8 steamroller cores on 28nm in an FX chip with a large cache and without an APU part. It would have been quite decent, but AMD could not afford the resources to make it and would probably be unable to sell it with adequate profit margins.

            • DPete27
            • 3 years ago

            The design was ahead of it’s time, that’s for sure. Unfortunately, like you said, AMD lost the gamble that the majority of software would just magically change to highly multi-threaded over night.
            Someone in their R&D department forgot that back when they launched their first multi core CPU, the processor market was all single core, so the revolutionary change was an easier feat. Nowadays, all CPUs are already multi-core, and Intel’s Bloomfield 4C/8T CPUs already existed when Bulldozer launched, so expecting software to just ADD thread requirements is not gonna happen as quickly as they needed it to. Hence the clock bumps to compete on IPC and resulting dismal perf/watt.

          • ronch
          • 3 years ago

          But on the bright side, when AMD copied Intel’s bad idea they surpassed Intel and made it even worse. Good job, AMD.

        • ronch
        • 3 years ago

        The only similarity between Netburst and Bulldozer is both are/were speed demon designs. Speed demon designs are not exclusive to Netburst so AMD probably didn’t seek to copy Intel’s missteps. They were most likely under the guidance of experienced IBM guys. Phil Hester probably had no small part in laying the groundwork for BD. Also, Bulldozer did introduce some novel ideas of its own, particularly the shared resources concept, despite the overall design failing to beat what Intel had. Nonetheless, it’s still amazing how a much smaller AMD managed to pull off a sophisticated design such as Bulldozer, and even more amazing how they have completed Zen, a far more ambitious design at a time when they had less money for R&D in the bank than at any recent time.

      • Kretschmer
      • 3 years ago

      Cool story bro.

      • the
      • 3 years ago

      Bulldozer was one of those architectures that one paper had some great ideas but as we all know, the road to hell is paved with good intentions. Sharing execution resources on both the front and back ends is a novel idea and had some great potentials in performance, die size and energy consumption.

      The flaws in implementation boiled down to four things:

      1) Cache topology. The cache design was write though which reduces bandwidth in favor of consistency. To further compound the cache topology, AMD continued to use exclusive caches.

      2) Decoder bottleneck. While the general idea of sharing decoders is a good one, AMD simple didn’t provide enough of them. Streamroller addressed this by providing dedicated decoder per threadAn alternative would have been to implement a micro-op cache to reduce the pressure on the existing number of decoders in the Bulldozer design

      3) Not enough floating point execution resources to compete against Intel. Bulldozers FP wasn’t bad by itself. Problem was that it was pretty much by itself. Intel was busy implementing dual 256 bit units vs. dual 128 bit units in Bulldozer. AMD should have incorporated four 128 bit units for flexibility and not fall behind in throughput.

      4) Designed for high clock speeds that it never hit. The same problem Prescott had and the same subsequent power guzzling result. It didn’t help that Bulldozer designs started at 32 nm SOI and never went past 28 nm bulk.

      • just brew it!
      • 3 years ago

      Rather than throw money down a hole, why not just have your existing FX-8350 mounted and framed when you finally retire the system? I think that would be more interesting from a memorabilia standpoint.

      • Concupiscence
      • 3 years ago

      You may be better served by finding the fancypants aluminum box earlier FX chips came in and putting that on your mantle. I’ve got one and it’s a weirdly proud little memento for me.

      • tipoo
      • 3 years ago

      Legitreviews took a last look at the best FX before Zen, kind of cool. But man, it gets brutalized by a humble i3 most of the time.

      [url<]http://www.legitreviews.com/amd-fx-9590-8-core-cpu-review_190566[/url<]

      • ptsant
      • 3 years ago

      If you actually code to the metal, the potential of an FX8370 is quite interesting. There are 16 ALUs in there vs 12 on a Skylake, but the cache and branch prediction is certainly not as good as in the Intels.

        • tipoo
        • 3 years ago

        Eh, the ALUs were half the width though.

          • ptsant
          • 3 years ago

          You are referring to SIMD instructions, I suppose. The width of a x86-64 instruction is generally 64b. The individual _core_ is wider in Intel because it has more ALUs (3vs2), but most importantly, these ALUs have shorter latency, allowing them to finish the calculation in less cycles. If you manage to hide the latency, a Bulldozer core can retire close to 2/3 the ALU instructions of an intel core…

    • derFunkenstein
    • 3 years ago

    If $3M is just missing, then I’m just barely short enough cash to buy a private jet.

      • UberGerbil
      • 3 years ago

      Well, it’s on $1.1 Billion in revenue. On a personal level, that’s like if you had a $50K salary and you ended up, what, $136 short at the end of the year after paying all your expenses (including taxes)? The real issue is that if this was personal, you’d be that guy who had taken out a bunch of loans in the past that were going to start coming due very soon.

        • derFunkenstein
        • 3 years ago

        Yeah, I guess that’s probably fair. Thankfully (aside from the mortgage) I am not that guy. 😀

          • UberGerbil
          • 3 years ago

          Also, if you were AMD Guy, you’d have sold your car several years ago, now reliant on the GloFloBus and TSMCUber to take you where you need to go. And your house as well, so you would now be living month-to-month in a crappy apartment, while eyeing that big empty cardboard ARM box in the alley as your long term backup plan.

            • derFunkenstein
            • 3 years ago

            That’s pretty hysterical. Nicely done.

      • ronch
      • 3 years ago

      Not sure $3M is enough to buy a jet, even one that’s Certified Pre-Owned, as second-hand stuff is nicely labeled these days.

        • Liron
        • 3 years ago

        There’s a 1998 Cessna up for grabs for $1,295,000
        [url<]http://www.theprivatejetcompany.com/private-jet-listings/id/205[/url<]

        • ptsant
        • 3 years ago

        A Cessna Citation Mustang costs, new, less than 3M. I would go for the HondaJet at 4.5M.

        I wonder, does AMD have a corporate jet?

          • ronch
          • 3 years ago

          No, but they have golden parachutes always ready to deploy.

      • AMDisDEC
      • 3 years ago

      $3M is nothing and not worth whining over.
      The Sales and Marketing teams spend more than this on meals and liquor in one quarter alone.

    • odizzido
    • 3 years ago

    Looking forward to seeing zen finally release. I am looking for an upgrade but intel hasn’t released anything that, to me, has been worth the cost.

      • DancinJack
      • 3 years ago

      So, AMD is going to have better CPU performance and, or, a better platform?

      That’s cute.

        • VincentHanna
        • 3 years ago

        If intel continues to sit around and wait for AMD to catch up, then it stands to reason that AMD WILL eventually catch up… or bankrupt themselves trying.

        • Anonymous Coward
        • 3 years ago

        Not clear that you are responding to what was written in the post.

          • DancinJack
          • 3 years ago

          Sure it is. I’m making that point that if Odizzido doesn’t think Intel has put out anything worthwhile, how on Earth does he/she think AMD’s offering is going to be worthwhile?

            • Anonymous Coward
            • 3 years ago

            Price.

    • Vaughn
    • 3 years ago

    This is great news a healthy AMD is good for us and even Nvidia.

      • Pwnstar
      • 3 years ago

      Yup.

    • Tristan
    • 3 years ago

    They do not need Ryzen to be profitable. Just let they throw CPU R&D, and limit itself only to make GPU for graphics cards and SoC for consoles. They may use ARM license for pennies to fill CPU in these SoC. Thsi way they can make easily 100-150 mln$ per quarter.

      • Anonymous Coward
      • 3 years ago

      Whatever possibilities exist in consoles and GPUs, the mainstream desktop CPU market and its fat margins are screaming for a challenger.

        • Tristan
        • 3 years ago

        Costs for design competitive CPU core is 3-4 bln $, and you need to sell dozens of milions of these CPUs just to break even, and twice of that to continue R&D (new revisions) and make some profits. Even Intel is skipping these next gen march costs, and constantly updates their Nehalem march.

          • Anonymous Coward
          • 3 years ago

          The R&D prices are dropping all the time, the market is hardly changing from one year to the next. ARM chewing on the low end, PPC chewing on the high end, AMD can just walk in and increase their share of the middle … even with an entirely vanilla product.

            • freebird
            • 3 years ago

            Exactly, not to mention all the forward thinking AMD has put into Zen and Vega… and their special sauce “infinity fabric” and HBCC. Could lead to dramatic latency reductions. Remember the FirePro with 1TB of storage?
            [url<]https://arstechnica.com/gadgets/2016/07/amd-radeon-pro-ssg-graphics-card-specs-price-release-date/[/url<] Just imagine when you have Zen + Vega + HBCC all on one die? If MS/Windows would steal/license some knowledge from IBM OS group on Single level storage it could help usher in more efficient OSes. Where DRAM, NAND, MRAM, HBM and traditional Disk storage are all just different "pools" of memory accessible by the CPU and other devices in one unified Storage Address Space. Sounds like fun times are ahead... just keeping my fingers crossed Windows can keep up.

      • crystall
      • 3 years ago

      Yes they do, and more importantly it needs to be a design that is not only competitive in the desktop/laptop market but also in the server one. Intel has been making money hand over fist in the enterprise segment since AMD essentially withdrew from that market due to the lack of a competitive core.

        • K-L-Waster
        • 3 years ago

        And it can’t just be competing on being good value for money. As numerous posters have noted, Intel has fat margins on their chips: if they start losing too much business to AMD based on price, Intel has loads of room to cut prices and still remain profitable.

        If RyZen isn’t legitimately competitive on pure performance and performance per watt, AMD is in trouble.

          • Anonymous Coward
          • 3 years ago

          Sure, Intel could undercut AMD’s prices, but they would do so by destroying their own margins. They won’t do it. They’ll lower prices only enough to maximize their profits, given whatever condition AMD is in.

            • K-L-Waster
            • 3 years ago

            Exactly — point being that if AMD starts taking enough business that it affects Intel’s sales significantly Intel will, as you say, reduce prices enough to maximize their own revenue. But if AMD’s primary selling feature is low price, as soon as Intel does that it will cut AMD’s sales.

            The only way AMD is in a good place is if RyZen is compelling for reasons over and above pricing.

            • Anonymous Coward
            • 3 years ago

            So what, this is how the entire world economy works. What matters AMD’s product is available in front of buyers at prices that make it worth while buying. They don’t need to beat Intel any more than Intel needs to sell only i7 models.

      • freebird
      • 3 years ago

      That is thinking way too small… AMD is thinking and shooting for the “Stars”!!! : ) i.e. Polaris, Vega, Navi…

    • chuckula
    • 3 years ago

    Su purports that RyZen is launching in March (not in February like that rumor stated).

    There was no firm date given though.

      • godforsaken
      • 3 years ago

      if I remember correctly, the rumor (assuming its the same one you’re speaking of) was that it would launch prior to march 3rd.. technically still allowing 3 days for that rumor to be correct…

      personally, when I heard the rumor my thoughts were that it would be the very beginning of march, so, this still fits into what I was expecting..

        • chuckula
        • 3 years ago

        You know what’s interesting?

        In November it was easy to find a full review of a retail 7700K just by throwing a rock on Google.

        The 7700K didn’t actually launch until January 3 of this year though.

        Now let’s think about RyZen.

        Let’s not think about the fact that roughly 1 month from its purported launch date that you won’t find similar highly detailed reviews of retail parts (not shady rumors about ES parts).

        Let’s think about the fact that a month from the purported launch nobody here can even name the model numbers for RyZen.

          • DancinJack
          • 3 years ago

          I don’t think that means much at all. There have been plenty of CPU releases in which next to no solid numbers/reviews came out until the day of “the launch.”

            • Voldenuit
            • 3 years ago

            Yeah, it was pretty much intel that started the early access ball rolling with Conroe and Sandy Bridge.

            Conroe was a big surprise after the Pentium 4, so intel really needed to impress. However, you could say the same of AMD right now. Maybe they just operate differently?

          • godforsaken
          • 3 years ago

          I honestly don’t care when it comes out too much, I wont be getting one (no need to upgrade right now)

          All I care about is it giving intel some competition because I don’t want another pentium 4

          • Ninjitsu
          • 3 years ago

          I’m assuming review cpus will reach the press two weeks before, with an NDA timed with the official announcement.

          • freebird
          • 3 years ago

          I’d rather think about all the money AMD will make while Intel has no comparable product that is price competitive in the 8-core and 6-core market over the rest of the year… not to mention how disruptive Zen will be in the server market… Ka-thump.

      • thedosbox
      • 3 years ago

      If you’re referring to the rumor sparked by AMD’s GDC presentation, that is scheduled for March 2nd.

      [url<]http://schedule.gdconf.com/session/optimizing-for-amd-ryzen-cpu-presented-by-amd[/url<] Some people may have got ahead of themselves because the conference as a whole is scheduled to run from Feb 27th - March 3rd.

      • Tristan
      • 3 years ago

      Let they postpone Ryzen launch to Q2, together with Vega at the same day. Zen is already almost year late (original plans was Q2 – Q3 2016), so quarter more or less do not matter much.

      • derFunkenstein
      • 3 years ago

      Remember when Lisa Su said things at the New Horizons event like “and while we’re saying Q1, Ryzen won’t launch at the END of the quarter”. Just a flat-out lie.

        • thedosbox
        • 3 years ago

        The GDC talk is at the beginning of March. They still have four weeks before the end of the quarter.

          • derFunkenstein
          • 3 years ago

          Depends on retail availability. I fully expect the “release” to be a paper launch. If that doesn’t happen until even March 1, I doubt it’ll be available in any real quantities by March 20th.

            • thedosbox
            • 3 years ago

            So what? This won’t be the first time a company has “soft launched” a product with limited supply.

            Granted, current owners of AMD CPU’s might be feeling the need for a refresh more keenly 😉

            • derFunkenstein
            • 3 years ago

            Just that you can’t listen to anything from the CEO if that’s how it pans out.

            • thedosbox
            • 3 years ago

            Once again Q1 closes at the end of March. If the product is available at retail even one day before the end of March, it counts as “before the end of the quarter”.

            Or do you have some arbitrary definition of “before” that nobody else uses?

            [edit] She reiterated the “early March” time frame on the conference call

            [url<]http://www.pcgamer.com/amd-confirms-ryzen-and-vega-launch-windows/[/url<]

            • freebird
            • 3 years ago

            That’s my biggest worry and if it performs too well, will they have to devote more to server production vs. consumer? Does sound like they are trying to still fine-tune production capability with market segments for Production release and how much they can ramp up production based on demand… capitalism is great, ain’t it? : )

            Depending on price I may go for a 4 or six core and upgrade later in the year as they get the process worked out better. I have a feeling they may price high due to high demand… especially since Server CPUs are supposed to come in Q2.

      • ronch
      • 3 years ago

      Yeah. It’s kinda nuts how AMD is being so vague here. Even info on the design of the retail packaging is nowhere to be found. You’d think they’re going full blast producing retail units by now, considering they said Ryzen will have a hard launch, but it seems there aren’t even any leaked photos of the retail packaging. What’s all the secrecy for? Or is this gonna be like a surprise birthday party?

        • freebird
        • 3 years ago

        I think AMD thinks Intel is underestimating how good Zen can be not only in the Desktop, but in the Server market as well. I think they want it under wraps until both are almost ready rather (aka March for Ryzen June/July for Server) than the 6-9 months that was previously projected with the Desktop and then Server chip launches. This let’s AMD fine-tune their Desktop launch and hopefully have a boat loads of chips for the server launch as well.

        As you can see from some of the articles, looks like the desktop motherboards are out there and ready to go from all the major suppliers. I’m kinda surprised no one on that side of the aisle hasn’t spilled the beans on how many motherboards they think they will need for the “launch date”?

        I think they want to launch a “broadside” into kabylake/cannon lake; because a six-core mainstream Intel chip won’t come along until coffee lake in 2018 supposedly.

        [url<]http://www.zdnet.com/article/intel-taking-its-six-core-processors-mainstream-in-2018-with-coffee-lake-family/[/url<]

Pin It on Pinterest

Share This