The leader-beans from Intel’s Executive Suite gave a conference call including some updates to guidance and results from Q4-2012. Mr. Market was not impressed and INTC’s stock price got taken to the woodshed, down some 6%.I’m not going to bore you with details about projections and financial blah blah blah. But, I am going to discuss why I think Mr. Market is mispricing INTC in an awful way and why I bought (and continue to hold) chunks of INTC (albeit at lower price points than the current per stub price).Everybody knows that INTC makes those chips that power your laptop, desktop, and server. And INTC has that market sewn up tight. Some 80%+ of all chips in traditional computers are made by Intel (I’ve even seen numbers as high as 95%). A tightly sewn-up market is great because it provides pricing power. And INTC obviously has that kind of power in spades: INTC has been able to generate operating margins north of 55% for quite some time in this “traditional” line of business (and some times north of 60%).Typically that kind of (marginal) profitability attracts competitors. But the “traditional” chip market is in decline to the tune of 3-5% per year. And one of the things I’ve learned from watching Buffett over the years is that industries in decline typically have a huge degree of pricing power (i.e. moat) because there is no threat from entry. In the case of traditional chip-manufacturing this makes perfect sense: who in their right mind would burn >$5 billion to build a chip fab, just for the “honor” of selling x86 chips into a market that is shrinking 5% per year? You’d have to be nuts! (not to mention crazy because INTC would build a more modern fab that made yours worthless in less than 24 months!)So, this fact alone might justify INTCs low-ish P/E of ~10 (trailing twelve months). While I think INTC shares are low-to-fairly valued for just the x86 side of the business, I think Mr. Market has completely ignored Intel’s planned (and announced) ramp into the mobile processor space . To see how INTC is accelerating their push into mobile you first have to understand how their simple R&D dynamic works. It’s all about “tick-tock.” Intel develops chip fabrication plants based on (duh!) various fabrication technologies. Their current cutting edge fabs are 14nm process and are only used to make (about half of) their “Haswell” line of “traditional” processors. Nothing mobile here. Yet. INTC also has fabs that make chips using 22, 32, and 45 nm processes. Haswell chips are currently made in both the 14nm and the 22nm process fabs. But, in 12 months ALL Haswell chips will be made in 14nm fabs and only in 14nm fabs. Then, in another 12 months the cutting-edge Haswell chips will be partly made in a 10nm fab (that is just now being constructed) while the 14nm fabs will continue to make Haswell chips but will also produce less “cutting edge” processors (i.e. the “Core” processor families). Currently, the “Core” processor families are made across the 22, 32, and 45 nm fabs. But in 12 months the “Core” processors will move up in process and will only be made in the 32 and 22 nm fabs. Then in another 12 months they’ll be made in the 32, 22, and 14nm fabs. You get the idea.Ok, so where are Intel’s mobile processers in this [cough] process? INTCs mobile processors are currently being made in the 32 nm process fabs. But, INTC has recently stated that their mobile processors are no longer on the “tick-tock” schedule. Mobile processors are now on a time-bomb schedule: “tick-tick-tick.” That means in three short years INTCs mobile processors will be made in their cutting-edge fabs on the same “tick-tock” schedule as their cutting-edge “Haswell” processors. So, uh, who cares? Well to answer that you have to consider the 800 lbs. gorilla in the mobile space: Qualcomm. QCOM is a so-called “fabless” chip manufacturer because they design their chips but then contract with chip foundries (manufacturers) like Taiwan Semi. and Samsung to actually make their chips. So, QCOM is at the mercy of those foundries’ manufacturing technologies. The cutting-edge foundries at Samsung and TSMC currently produce chips using a 28nm process (note that this is double the size of INTCs current cutting-edge 14nm fab but smaller than INTC’s current mobile process size of 32 nm). This year INTC is putting out 22nm mobile chips (this is the first “tick” of the time-bomb schedule). While INTC will have this “better manufacturing tech” they should be handicapped because they don’t have the same depth of experience making mobile chips. So let’s call INTC and QCOM a “tie” for their 2013 generation of chips. But in about 12 months INTC will start pumping out mobile chips with their (already-in-use-by-the-Haswell-family) 14nm fabs. Meanwhile QCOM will be stuck with the 28nm TSMC and Samsung fabs putting INTC about 12 months in front of QCOM. (that’s the second “tick” of the INTC time-bomb).But wait there’s more. The third “tick” of INTCs time-bomb is that after their mobile processors spend 12 months with a 12 month lead on the QCOM chips they will then tick their way into the 10nm process fabs at the bleeding edge of INTCs manufacturing processes. That’s a 24 month lead on QCOMs chips (chips that will be made in whatever ~20nm foundry TSMC and Samsung have built by then). Intel goes BOOM!Ok, so this is a complicated story. And it’s no guarantee that INTC will end owning the mobile chip market like they do the “traditional” chip market. Mobile processors are partly about integrating the main processor (CPU), graphical processor (GPU), and cellular/wi-fi bits (these so-called “Systems on a Chip” or SoCs). And so far INTC has yet to demonstrate their competence with SoCs by a “big-win” in the mobile processor space. But that will likely change over the next 6-12 months as their current mobile processors are “tied” with QCOMs best offerings but then jump ahead in the next 12 months. And while it’s certainly possible that QCOM is going to be able to compete against INTCs SoCs on an efficiency basis (even though QCOM will be handicapped by the 28nm processes of TSMC and Samsung relative to INTCs 14nm SOCs). But, even if QCOM can beat out INTCs 14nm SOCs on efficiency, there is no way they will be cheaper. And that “cheapness” margin will get even bigger when INTCs SoCs get made on a 450mm wafer in their 10nm fab.A final parting shot. If Intel decides to compete in something you do (that’s within INTCs circle of competence) you should be scared. If history is supposed to rhyme, then I’m sure any one of us can use iambic pentameter to tell the sad, sad saga of Cyrix and AMD.
Nice article but it misses a few key points. The manufacturing process is important if and only if you have the product your clients want. Intel is linked 100% to the PC world, that's why it's called Wintel. That world is hitting a wall as PC sales stagnate and customers move to mobile computing. Mobile is different. You need real time wireless stacks, you need low power, you need to be physically small as real estate is expensive. This is the world of ARM not x86. That's Intel challenge. And at the end comparing intel to the X86 clones like AMD is not justified. In the early PC days Intel had a monopoly. Along came AMD and reversed engineered the Intel chips and tried to compete on price. Never going to win more than a few % points.Now Intel are being forced to compete with QCOM in the mobile space with an inferior product. It will be tough for them. Having a great fab is a start but not the key to the castle.Tango on an IPad - that's intel free.
tangoev said: Now Intel are being forced to compete with QCOM in the mobile space with an inferior product. It will be tough for them. Having a great fab is a start but not the key to the castle.Not exactly sure what you mean by "inferior." But, according to AnandTech, some of Intel's current digs beat out even AAPLs über-efficient ARM-based designs, at least in the tablet space:http://www.anandtech.com/show/6529/busting-the-x86-power-myt...The phone to keep an eye on is the Motorola RAZR i. It's INTCs first win in the mobile & phone space. So far it's reviewing well for what it is.If INTC makes chips that will cost less and have specs that run circles around even AAPLs best ARMs then I don't care whether there is software for x86 tablets already or not. The hardware will get built because the hardware manufacturers will be able to smoke the iPad's specs. They will be able to make a better piece of hardware for cheaper. And the software will follow.My read on this situation is that INTC has been driving with only four of their cylinders running and yet they're knocking on the door of state-of-the-art. When they hit the accelerator (and they've said they just put their foot down), the other four cylinders are going to engage. If I were QCOM I'd be losing sleep because I only brought my buddies' 4 cylinder engine to the drag race. INTC is going to be back in the pits, eating QCOM's lunch before QCOM even gets done with the 1/4 mile.
Sorry, forgot to caveat that the INTC based tablets have much lower resolution screens. So, it's not exactly an AAPL to AAPL comparison!
Meanwhile QCOM will be stuck with the 28nm TSMC and Samsung fabs putting INTC about 12 months in front of QCOM.Just curious, why do you think that Samsung and TSMC will not build their own 22 , 14, 10 nm fabs. Are these technologies proprietary to and patented by Intel? Even if Intel is currently in the lead with the newer fabs, that is only likely to be temporary.And for customers who crave iPhones and iPads they are most unlikely to switch to another device by non longer cool Motorola, just because it has a chip inside made more cheaply on a newer fab. I don't think most Apple fans know or care what chip is inside their device.Unless, you contend that all these newer fabs will allow Intel to woo Apple away from ARM & QCOM to its own designs. That seems like a far- fetched assumption to me.
Not to mention that current ARM phones don't have the latest ARM design.
There's nothing special about INTCs fabs...TSMC or Samsung could build their own.But, INTC already has a 14nm fab that is currently making chips. TSMC et al., as I understand it, just have their 28nm foundries and have not broken ground on any new fabs. In contrast, INTC is already building their 10nm fab that will use 450mm wafers. That will make their chips really efficient and really cheap.It's certainly possible that TSMC or Samsung will just go right to a 10nm-ish fab. But, again as I understand the situation, foundry margins aren't good enough for them to be on the bleeding edge of chip manufacturing...they're usually 1.0-1.5 steps behind the most current stuff. Perhaps AAPL can use that slug of cash to pay for a fancy fab for TSMC (unlikely that they'd pay for Samsung). Or perhaps INTC would be willing to act as a foundry for AAPL. Doubtful. But possible. Also, it's not like these fabs come out of thin air. New tech has to be developed at each stage to make them work. And INTC is leading the charge with their 14nm fabs that use some fancy 3D transistors... In terms of the specific smartphones, I'm dubious whether AAPL would go to an x86 mobile chip made by INTC unless it really started crushing their own ARM designs. Moreover, I think INTC is targeting QCOMs current domination of the Android and Windows based phone and tablet space.
What's so great about a 14 nm process for mobile chips?In traditional computing of course, smaller is better. A 14 nm process can fit 4 times as much logic in the same area as a 28 nm process.But the big big big cost is leakage current. Shrinking process geometry is not a free lunch. Leakage current means you lose power efficiency very quickly. A naive shrinkage (which I am sure Intel is not doing) would convert your processor into a nuclear reactor. Again, this is a big problem in a traditional PC, but you can still have massive amount of cooling and sort of get around it. You cannot fir a heatsink and a fan in your cellphone.Intel's Achilles' heel is power. Arm chips were designed from the start to be low-power, x86 were not. And now Intel is paying for it. 32 nm and higher geometries still allow your tabl-cellph-all-in-one-device thingie to have plenty of processing power. I won't do a Bill Gates here and say 640K is enough for everyone, but the point is that 32 nm it's not exactly horse-and-buggy either. Power matters much more than geometry and Intel hasn't demonstrated their prowess yet. Haswell is promising at 22 nm and 10W (lowest power, for tablets), no doubt. But it's not a done deal.
As you said knighttof3, a jump to a smaller process is no guarantee of success. But it looks like the TSMC et al. foundries are way behind on the state of the art FinFET processes due to a tactical gamble that went the wrong way for them:http://www.extremetech.com/extreme/142926-intel-unveils-22nm...Also, my impression is that QCOM's lead in the power-efficiency space is based primarily on the fact that they developed an out-of-order processor while Intel's super-efficient processor (Atom) is an in-order processor. As I understand it, Atom is just an old Pentium design that has been scaled down. So, effectively there is already a roadmap in place for the circuit deployment if they decide to stick with in-order. But, even with Intel's in-order handicap they're still competitive with the OoO designs merely because their manufacturing is better than everybody else's. Word on the street is that INTC is developing an OoO Atom. That should begin the erosion process for the one place that QCOM has an advantage. I still think INTC could pull ahead in efficiency with their in-order designs based purely on their fabrication prowess. And if they build a decent OoO processor then INTC chips will likely run circles around the QCOM offerings.It's basically a story of "Intel now 'gets' mobile." (Moreover, mobile is now big enough for INTC to pay attention to). If you're QCOM this should keep you up at night because your foundries just did a stutter-step on their manufacturing tech. while INTC just took two strides.At the end of the day we're trying to buy mis-priced risk. And I think INTC is priced reasonably for just its traditional line of business. But, INTCs mobile side is like a lottery ticket with good odds of hitting (at this point in time anyway). Of course a "win" in this space could end up making no profits and the leader-beans could certainly bleed the profits from the traditional lines into a boondoggle. But there is that 4% dividend and a deep management team that knows how to allocate capital. So I'm less concerned about that. My guess is that in five years INTC will be in the #2 or #1 spot in terms of mobile processor sales (let's exclude AAPL here since they just roll their own). And they're going to do that by eating QCOMs lunch. But, right now, INTC is being priced as if QCOM has already stolen it's milk money.
Inferior - I used this as judged by the market at present. Vast majority of mobile devices are not Intel powered. The vendors chose the best CPU for their needs: This is always a compromise of cost, power, size and importantly developmental tools and software.One other point is that the mobile market is expanding at a huge rate. Intel has to gain market share as the PC world is stagnating. It will be a changeling time for them.Tango
* Intel's desktop/server processors cost > 10 times ARM & it's own ATOM processors. This has two implications: i) on the server/desktop side there is a big incentive for ARM vendors to take _any_ market share. If QCOM/NVDA/MRVL/BRCM etc. can gain even 5% share at half the price of INTC chips their revenue and earnings will explode. So far they have tried through the backdoor of windows 8. I suspect they may have better luck in servers. ii) because of the difference in revenue per chip, for every $10 ATOM intel sells at their cutting edge process, they lose the sale of a $200-$1000 core or xeon. This, combined with the fewer processors (due to low yield) they can manufacture on newer process, ensures that the mobile processors are not on their newest process in volume.* Intel had the advantage of exploiting something called ILP (techniques developed by mainframe vendors in 60s/70s) to ensure that each generation of processors would be significantly speedier then the prior. This advantage went away a decade ago. ARM has this advantage now of putting in these enhancements while Intel is unable to push the performance envelope very fast. In processors 1/10th the cost.* The mobile "processor" is different then the desktop processor. The mobile System on Chip (SoC) has many processors in it, only one of which is the "CPU" as understood on the desktop. There are DSPs for vocoders, specialized processors for dealing with peripherals and a set of processors for wifi and modem. This involves a great deal of system integration and therefore a solution (software and hardware) that is proven over many years and product cycles will be easier and quicker to bring to market.* Power consumption is fundamentally about turning off circuits when they are not needed. This is easier for a simpler and specialized processor like the ARM then the general purpose x86. Think of it as power consumed by a small house versus a large casino. The house consumes less power to begin with and when it comes time to turn off the lights it is easier to turn off the few lights in a house.Intel's core market will be (and already is) under assault. The market size and margins are too lucrative.In the meantime Intel is competing from a position of weakness in the mobile space. It is competing for a small sliver of the market given the large Apple and Samsung markets (50% in phone, 80% in tablets) are out of reach to it. Further it cannot bring to bear its tremendous advantage in fab technology because this takes away from its much more lucrative "big" processors.It will be a fascinating battle. I don't believe it is advantage Intel currently.
Don't forget about Nvidia chips, their graphic chips have been showing good performance inside of servers (with less power). It is early and a small threat but it may be a threat that is building for Intel.
Intel could be bought just for the over 4% dividend yieldAnyone know how they missed the mobile market? Why weren't they working on this a few years ago with chips already to go?
Why weren't they working on this a few years ago with chips already to go?But they were with StrongARM and then with xScale (both ARM based) which they sold to Marvell Technology Group Ltd. to switch to x86 based Atom.XScaleOn June 27, 2006, the sale of Intel's XScale assets was announced. Intel agreed to sell the XScale business to Marvell for an estimated USD 600 million in cash and the assumption of unspecified liabilities. The acquisition was completed on November 9, 2006.http://en.wikipedia.org/wiki/Marvell_Technology_Group#XScale...http://www.marvell.com/Denny Schlesinger
But they were with StrongARM and then with xScale (both ARM based) which they sold to Marvell Technology Group Ltd. to switch to x86 based Atom.And ironically enough, AMD did something similar in 2008, selling off their Imageon ARM IP to Qualcomm, who renamed it the Adreno (an anagram of Radeon).
>ii) because of the difference in revenue per chip, for every $10 ATOM >intel sells at their cutting edge process, they lose the sale of a $200->$1000 core or xeon. This, combined with the fewer processors (due to low >yield) they can manufacture on newer process, ensures that the mobile ?> >processors are not on their newest process in volume.Some good points. I think this one is a little off though. Intel should be able to make 10 Atoms on the same area as a single Core i7. Are the margins going to be enough to make up the difference? Is there enough demand to sell that man atoms? Does "going pedal to the metal" include producing Mobile processors on 300mm wafers? That would drop their costs even more.The math:Core i7 @ 45nm is about 263mm^2, while atom & arm are running about 20-30mm^2 @ 40-45nm.So shouldn't Intel be able to get roughly 10 mobile processors for the same real estate as a server processor? Also yields will be higher since a single error is killing a much smaller functional patch of the wafer now. -Dan From memory, ARM/Atom cpu's are
Some good points. I think this one is a little off though. Intel should be able to make 10 Atoms on the same area as a single Core i7. Are the margins going to be enough to make up the difference? Is there enough demand to sell that man atoms?Does "going pedal to the metal" include producing Mobile processors on 300mm wafers? That would drop their costs even more.The math:Core i7 @ 45nm is about 263mm^2, while atom & arm are running about 20-30mm^2 @ 40-45nm.Thank you for sharing your thoughts, Dan. The atom Intel makes for mobile is not a standard part but an SOC with integrated GPU. Intel does not give die size (or cost) for this part. Do you know definitely that this is 20 mm^2? I will be very interested.As a reference the A6 die size is 100 mm^2 at 45 nm and it, like the Atom, does not include the LTE modem & wifi so far.
Thanks for catching my mistake. I did use the desktop Atom die size.Apple's A6 probably isn't a bad place to start an estimate for Intel's die size. Though according to wikipedia the A6 is @ 32nm, so for apples-to-apples (no pun) we should use something like Intel's 32nm Sandy Bridge 4c w/ 4 cores @ 216 mm^2. 216 vs 100, so even without the radio, it's more like 1.5 to 2.5 Atom SoCs for the same wafer cost a single 4 core desktop chip. Also, if they add the radio on-chip, I'd guess it would add 25% or more area. Which drops us closer to 1.5 to 1 ratio. (Based on comparing Tegra4's outboard digital radio to the SoC size.) So the die size considerations I raised earlier don't have much weight with more realistic intel die sizes. It's going to be fun watching this unfold.-Dan
It's certainly possible that TSMC or Samsung will just go right to a 10nm-ish fab. But, again as I understand the situation, foundry margins aren't good enough for them to be on the bleeding edge of chip manufacturing OK. So I understand you correctly, what you are saying is foundry margins are poor hence TSMC and Samsung (?!) cannot afford it. However INTC can because they make high margin on x86, that should allow them to buid these 10's of Billion Fab's.Just wondering, if INTC becomes foundry for AAPL and others, what it will do to its margin? and how will market handle the margin erosion? What kind of valuation adjustments investors has to make?While the scenario you pained is interesting, it is not an easy choice to get into foundry business big time.
Best Of |
Favorites & Replies |
Start a New Board |
My Fool |