Intel (INTC) has been recently trying to catch up in the mobile space. They have revamped their mobile CPU lineup to place added emphasis on performance/watt, and are using their process advantage over the competition to try and finally enter significantly into the mobile market.
But there are a few things that have made me hesitant as to whether they can catch up, including how these factors will affect revenues and margins, and how many units they will need to move to show substantial growth.
Why Pricing is Tricky
Image taken from IBTimes.com
So the price of apps processors above range from $30 for an 8 core Samsung (SSNLF.PK) processor to $17.50 for a quad core. Companies like MediaTek probably charge less, but make up for it in volumes; Intel will be competing in a market sector where $50 is higher than the high end.
This presents several "red flags" to me.
Intel Inside Smartphones
Two articles on SemiWiki.com (I, II) were written describing smart phone sells in China, and how companies are able to produce these phones that retail at prices of $50-$75 for the entire phone (and I believe these are not subsidized prices).
The following tables taken from IDC.com
First I would like to point out that between 2009 and 2010, Apple (AAPL) and Samsung demonstrated 89% and 318% respectively, compared to 47% and 129% between 2011 and 2012. Also, Samsung's growth seems to primarily come from stealing market share from the other major players excluding Apple.
The "others" category, the place where the cheaper smartphones dwell, has seen sustained growth, and these "others" combined ship more smart phones than either Apple or Samsung.
Given that these smart phones are considerably cheaper than what most of us will find in our Verizon store, I assume that Intel will not be participating in this category, leaving them mainly in play for the higher end smart phones.
Intel Inside Tablets
You can see above that Apple owns the lion's share of the tablet market, followed by Samsung, Amazon (AMZN), and Asus. Note that the developing countries are seeing an explosion of cheaper tablets as well, but again this should not matter to Intel, since they are not in play in these cheaper markets in my mind. Amazon also aims to price their tablets very cheap, and Apple designs their own chips.
I will assume that Asus, Samsung, and "others" such as Acer and Sony (SNE) are the only sockets up for grabs from Intel.
What Happens if Intel Steals Considerable Market Share?
Given that it appears $30-35 is the high end of smart phone and tablet processor range, excluding the expensive Surface Pro tablets, let's assume that Intel's processors are so much better they convince OEMs they're worth an ASP of $50, which represents a 67% premium over the competition at a high end of $30.
Now let's also assume that Intel is able to grab all the sockets from Samsung smart phones and tablets. That's 215M smart phones, and roughly 25M tablets, for a total of ~240M devices.
This represents an increase in revenues of ~$12B each year.
Given their $53B in revenue from last year, this represents a ~22% increase, if Intel is able to ship 230M devices next year, and charge by my estimates a premium over the competition. Most likely this will not be an overnight transition, and it will take a few years for this to unfold during a time when the PC market is stagnating; Intel derives a majority of their revenue from the PC market.
Given that roughly 850M tablets and smart phones combined shipped last year according to IDC data, Intel would have to gain about 30% of the mobile market for this scenario to happen. With no high volume design wins announced yet, I find this highly unlikely. If Intel competes on price to gain market share, I feel they have to worry about lower GM.
Gross Margins in Mobile
SA contributor Ashraf Eassa published an article predicting margins on Intel's mobile chips. In the article he estimated Intel's Core CPU line selling for an ASP of ~$120. As for a second data point, Intel's mobile Core i3 ULV CPU is $287 per the ARK website. A discount of 50% over the ARK website places OEM bulk pricing at ~$145, so I will use $130 as an estimate for calculations based on the ASP of Core CPUs.
For the time being, Intel is using 300mm wafers with plans to transition to 450mm. A 300 mm wafer has roughly 70,000 mm^2 of usable surface area. I will present a range of values that I believe should encompass where Bay Trail's revenue per wafer will fall within by performing a few sets of calculations.
Intel's ULV mobile CPUs are 2 core, and come in an array of different GPU sizes. Here, I will assume an average die size of 170mm^2, which is slightly smaller than the 2 core chips with the larger GPU die area. As for Bay Trail, Nvidia's (NVDA) Tegra 4 is ~80mm^2, while Apple's A6X is ~120mm^2. Therefore, for Bay Trail die size estimates I will assume a die size between 80 mm^2 and 100 mm^2.
Regarding pricing, I will assume Bay Trail has an ASP of $30-$50 for OEM pricing. As die size goes down, yields tend to improve, so for ease of calculation, I will assume that Bay Trail yields at 100%, while Haswell yields at 80%.
So for Haswell, using the assumptions above, the revenue per wafer should be ~$43,000 (70k mm^2/170 mm^2 * 80% * $130). Bay Trail should fall between ~$24,500 (70k mm^2/100 mm^2 * 100% * $35) and $43,500 (70k mm^2/80 mm^2 * 100% * $50).
I know this is a pretty broad range, but these calculations are meant only as demonstration purposes since most of the information needed to make them accurate would be information that only Intel engineers and management would have access to.
So in my best case scenario assuming a die size of 80 mm^2 and a $50 ASP, Intel will generate revenues per wafer ~equal to their Core CPUs. In my worst case, Intel is looking at ~60% of the revenue per wafer of Bay Trail compared to Haswell. And this is also assuming Bay Trail yields 20% better than Haswell. GM is calculated using revenue and cost of sales. I am neglecting the increase in cost of sales from packaging, binning and testing a larger number of chips per wafer for Bay Trail.
In my best case scenario, it seems to me that Intel will be stretching it to generate an equal revenue per wafer when comparing Bay Trail to Haswell, so I am having a hard time coming to the conclusion that Intel will be able to maintain ~60% GM for their mobile products.
This in and of itself is not necessarily a problem provided Intel can sell Bay Trail in sufficient quantities. Assumptions that will affect the accuracy of my estimations are inaccuracies in the ASPs used, yield percentages, and die size estimates.
I have made the case before that the power consumption of the competition when compared against Intel needs to be taken relative to battery life and performance, since the end goal of focusing on performance/watt is to deliver the combination of maximum performance and time unplugged.
Intel's 22nm process has an inherent power savings advantage. But there are other ways chip makers can make architectural changes to their chips to save power. ARM Holdings (ARMH) uses a "big.LITTLE" arrangement to turn off larger cores and gives control to weaker cores, but this is the less interesting solution to me personally.
Qualcomm uses an asynchronous architecture. Rather than me trying to explain it, Qualcomm has provided a video that explains it better than any attempt I could make. These kinds of technology improvements have allowed Qualcomm to increase performance without dramatically increasing power consumption.
Motorola (MSI) recently released a new phone with an interesting setup as well. The solution, the X8, features a cheaper custom Qualcomm SoC combined with 2 separate DSP (digital signal processors): a "contextual computing processor" and a language processor. The crux of this is that the contextual processor can serve as the main processor in standby to lower power consumption of the Qualcomm's already frugal S4.
To assume that Intel will completely obliterate the competition in regards to power consumption would require looking at the mobile sector in a vacuum with your eyes closed, especially since we have not seen real world benchmarks that compare performance per watt, and we have no idea on pricing.
Although recently mobile benchmarks are coming under heavy scrutiny (I,II), the available GeekBench scores for the S800, S600, and Tegra 4 compared to the z3770 (the Tegra 4 is an unfair comparison though because it's taken from Nvidia's Shield which is actively cooled). I have also included a benchmark with the outgoing generation Atom to give a point of comparison for Intel competing in the K900.
Bear in mind, this is a single data point (GeekBench), and the most important performance/watt metrics are missing. There are potential issues with GeekBench in benchmarking x86 floating point performance as well, but for the time being with all available data, I am not seeing a definitive performance advantage for Intel. And regardless of what benchmarks say, I am waiting for real world tests to make a definitive conclusion.
I think to assume that Intel stock is going to take off and leave investors playing catch-up is analogous to saying that a turtle is getting ready to start sprinting. I have no opinion on Intel as a long term investment; I have been contemplating to adding Intel to my IRA and forgetting it's in there until I'm ready to go find a mountain lake somewhere to live out my days on. But for at least the remainder of 2013, I am not expecting much growth from Intel. And aside from some unforseen catalyst happening, I am waiting to see how well Bay Trail is received before I revisit my opinion.
As far as smart phones, Intel would have to capture significant market share while charging more than the competition (given iSuppli's price of $20 for a Snapdragon 600). I have shown why I feel that Intel competing on price will drive margins down which will limit upside in the stock. The three largest vendors for smart phones are "Others", Samsung, and Apple (in order). Intel is probably not competing in the "Others" category given this is where the cheaper smart phones reside. Apple I feel is out of the question until it's announced they're switching to Intel. Apple has spent a significant amount of resources in developing their own processor, and rumors report that at least for the time being, Apple is going to continue using their own in house chips. Samsung uses a mixture of chips across all their devices, so I do not see Intel capturing every Samsung socket. That was for demonstration only.
Major catalysts, like an announcement of Intel-powered iPads or iPhones from Apple, are the reason I keep cash on hand, so I can invest early and ride the momentum.
Regarding tablets, Apple also uses in-house designed chips, and iPad variants represent the majority of the tablet market. Amazon tablets compete on price, leaving Intel out of the question here in my mind as well. This leaves primarily Samsung, Asus, and possibly other OEMs such as Sony. However, given the total tablet shipments last year were roughly 130M, and that Apple and Amazon represent roughly 75% of this market, that leaves about 32.5M sockets for Intel to compete for.
And the competition is fierce. Qualcomm I feel is Intel's strongest competitor. They use a unique architecture that allows their processors to operate at lower power levels while providing good performance. Here, I will note that there are no independent reviews comparing Tegra 4, Snapdragon 800, and Bay Trail so nothing is definitive yet. I will revisit this conclusion if I see proof that Bay Trail is significantly ahead of the competition in performance/watt and competitive on price; I do not consider PR slides as proof.
Considering that Qualcomm is the same size as Intel by market cap, has a very wide and deep moat of IP regarding mobile processors and communications, and Qualcomm is accustom to competing in mobile where $35 is the ASP of a high end processor, Intel will face hurricane-force headwinds from a competitor that is equally sized and heavily entrenched in mobile.
If Intel is competing on price, I cannot see them maintaining GM, which I feel will limit the upside potential to the stock. Potential ways I could be wrong are if the die size of Bay Trail is considerably smaller than I have assumed, or if the performance is much greater than the competition so Intel is able to command a higher premium. R&D spending is accounted for between GM and operating margin, so Intel needs to maintain healthy gross margins to maintain a healthy R&D budget, or it starts eating into their bottom line.
The biggest advantage I see going forward for Intel is their x86 ISA and a cheap, competitive x86 solution in Bay Trail. OEMs need to put together competitive Windows based solutions not subsidized by "bloatware" that can compete on price and functionality with tablets, and not give off the vibe of a disposable computer.