I've been an unmitigated Intel (INTC) bull for the past two years. I have been wrong for those two years, yet I am playing the name bigger than ever for 2014. Why? Intel, as an enormous enterprise, takes much longer to do things than many, including me, understand. Intel engages in huge projects that take years of planning and design to bring to fruition.
Intel is criticized for:
- Buying McAfee
- Declining PC sales
- Missing mobile
- Overspending on Cap Ex
I have a theories that might explain these things one by one.
First, buying McAfee:
I think it is clear the Intel identified security, or lack of security, as the major problem in the path of all computing. Malware is an annoyance for consumers, but it is an enterprise threat for large companies and even governments. It is clear what Intel's intentions were. They told us clearly that they intended to combine the McAfee software with hardware feature (hooks) in the CPU chip to vastly improve the protection of all computing devices.
In a perfect world things could happen on your schedule; in reality, you sometimes have to move at non-opportune times. I don't think Intel was ready to deliver on the security promise when they bought McAfee.
At the time of the purchase of McAfee (August of 2010) the Sandy Bridge (SB) processor had not even been released. The SB i7 release came in January of 2011. Sandy Bridge only had rudimentary embedded security features. According to the Intel Tick/Tock strategy, the Sandy Bridge was a "Tock" (new architecture/stable process) and needed to be shrunk to the 22nm process and would be called the Ivy Bridge. Ivy Bridge was a "Tick" (stable architecture/new process). Ivy Bridge was introduced in April of 2012, just 15 months after Sandy Bridge was introduced. The earliest that the more robust security features would be available would be with the introduction of Haswell. Haswell (a "Tock", new architecture/stable process) was introduced in June of 2013, just 14 months after the introduction of Ivy Bridge.
Moore's Law states that the number of transistors on an integrated circuit would double about every two years. For a long time this two year cadence drove Intel new processor development. So the Tock to Tock time should be about two years. The Sandy Bridge to Haswell (Tock to Tock) has taken a little longer since the shrinks are just plain more difficult.
So, with that background, Intel knew when they bought McAfee that it would be 2.5-3 years until the promise of significantly improved security would be realized.
We now hear more about computing security from Intel. Here's a video that talks about the enhanced security features of the Haswell over the Sandy Bridge/Ivy Bridge architecture. During his keynote address for CES on January 6 2014, Intel CEO Brian Krzanich talked about FREE security for mobile devices (beginning at minute 28). This free security would not be limited to Intel architecture, Brian mentioned it would work on ARM devices as well…mentioning ARM by name.
Now for the stuff that was not said: While the security package would be free for non-Intel devices, non-Intel devices wouldn't have the hardware hooks that would make the security protection much more robust. Only Intel devices would have that sweet little feature.
So, Mr. Krzanich just subtly reached over the device manufacturers heads to the end consumer who, if he wants the best security protection, would arrange to buy an Intel powered mobile device.
The McAfee software overlay for Haswell, and future CPUs, could easily add 20-30% of Haswell revenue ($8-12 billion) CUMULATIVE over the next five years, resulting in an ultimate, and recurring, software revenue number that is the same as CPU revenue, but with 80-90% gross margins. The highly improved security protection of the processor hooks COMBINED with McAfee software could also be the prime driver of the Intel mobile business. Does buying McAfee still look dumb?
Let's tie this to the slowdown in the PC business.
Some individuals and every single corporate IT guy in the world understood what Intel was saying; computing security will get much better when Haswell gets here to mate with McAfee. Any corporate IT guy would have to be crazy to allow a comprehensive PC update until Haswell machines were available in the format he needed. So we have something of a partial Osborne Effect at work on the PC business since the purchase of McAfee. My take is the PC business is not so much in decline as in "delay". Any further explanation by Intel on the security issue would only exacerbate the Osborne Effect.
With the mass availability of Haswell based PCs, expect Intel to begin advertising their advantage in security, which will stimulate the corporate refresh cycle and individual upgrading of a very old fleet of PCs.
The third party PC sales reports generally come in the week before the week of Intel earnings, except in the first quarter. The report of fourth quarter PC sales seems to come in when CES is over. If that is true we should know by Monday (January 13, 2014) what PC sales were in the fourth quarter and some projection on where they will be going in 2014. There is some indication that PC sales have bottomed and might even be going up. There are also many more reports of a continuing decline in PC sales. These reports seem to be regurgitation of the same source material.
If PC sales show even no decline and indications of a future increase, Intel will have a very interesting earnings week.
OK, let's talk about "missing mobile".
in 2011, Intel sales spiked up about 25%, most of which was new business in the data center segment. So while the world was distracted by the Great Recession and the growing smartphone phenomenon, Intel was busy finishing up on the 10 year effort to dominate the server chip and high performance computing chip business. Intel did miss the mobile business because they were engrossed in developing and providing the very technical infrastructure that made the mobile business possible.
The Data Center business turned out to be another $10 billion business for Intel at 90% gross margin and essentially no competition. So they didn't really miss the mobile business, they simply chose the high end of the mobile business. The emphasis on performance in the server business fit much better with the Intel capabilities of the time. Data Center over mobile was the right decision for Intel and for the mobile business as a whole.
As we might see from the security comments above, Intel will very likely use their unique position in hardware/software based security to drive a wedge into the mobile device business starting this year.
The most logical mobile starting point for this surge is in the tablet segment. The tablet segment can still use the higher performance of Bay Trail and has room to use a discrete LTE chip and then in only in some percentage of tablets. Intel has the required LTE baseband function in the XMM 7160.
An LTE baseband function integrated on the Application Processor (AP) SoC will be a little further out.
It might turn out that a monolithic AP+LTE is not the lowest cost solution. According to former CEO Paul Otellini a multichip package might be the better solution due to yield issues. He did say that the LTE needed to be in the same package as the AP for power management optimization.
This seems to be possibly borne out by an analysis of Qualcomm's (QCOM) operating margins on the IC division. Qualcomm only generates about 16% operating margin (page 7) on the chip division. Compared to similar fabless semiconductor companies, such as Xilinx (XLNX) and Altera (ALTR) at 27% and 33% respectively, the Qualcomm operating margin is low. Something is eating into profitability at Qualcomm and it is probably bad yields on the integrated AP and LTE.
So, LTE, with the XMM7160 and the soon to be released XMM7260 will cease to be an impediment to Intel's growth in the mobile business.
The mobile summary is that 2014 will be the year of Intel in mobile driven by the elimination of LTE impediments and a superior and free security solution that can't be matched by any other supplier in the mobile chip business.
Overspending on Cap Ex
Intel has spent a total of $45 billion over the past four years on manufacturing facilities. Maybe this capacity is needed for serving the mobile business, but the entire mobile business couldn't fill the extra capacity being finished by Intel.
I've suggested that the massive capacity could be for an attack on NAND memory for use only in solid state drives (SSD). I often feel lonely in that opinion, but here we go again:
Intel has made materials breakthroughs since 130nm in order to keep Moore's Law (ML) progressing. Intel has brought Strained Silicon to semiconductors when everyone thought that ML stopped at 130nm. They brought HKMG when everyone thought that ML stopped, for sure, at 65nm. They enhanced HKMG for 32nm, and brought TriGate HKMG at 22nm when everyone else thought ML CERTAINLY had to be finished by 32/28nm. Now 14nm TriGate is spitting distance away from being in your notebook or tablet or smartphone.
Intel is probably the closest thing America has to the Bell Labs of the 1940s.
From a business standpoint, Intel is only number six in wafer capacity. A good question could be, "How long can Intel allow Samsung to produce twice as many wafers as Intel?" My opinion is that every wafer produced carries with it some molecule of LEARNING that in aggregate, over time, makes the highest volume producer also the technology leader.
The new Intel fabs, whether 300mm or 450mm, represent about 600,000 Intel 200mm equivalent wafer starts per month of capacity NOT listed in the article (the article, by the way, is listed in 200mm wafer equivalents). If Intel, through another materials breakthrough, were to burst on the scene with a good 10nm NAND chip at 256Gb, that, at 10nm, would be 100sq. mm in size, it would produce about 550 die per 300mm wafer (adjust that up for 450mm wafers) and cost about $3.60 per chip, or about $.11 per GB of storage. That is close enough to HDD prices to put them out of business.
If SSD storage had to be $.30 per GB to cause a virtual explosion in growth of SSD, Intel could do it with such a part and generate 60% margins in the process. The very best COST from any NAND producer today is about $.40 per GB. To make even 30% GMs these producers would have to charge about $.60 per GB...which they are. $.30 per GB for Intel and $.60 per GB for all others. Who do you suppose would win that competition?
The market for HDD at $.10/GB is about $40 billion. A large part of that HDD storage goes unused in the client business; who can use 500GB let alone a TB in a casual computing environment?
What that says is that the SSD business could deliver about half of the HDD GB and still satisfy the storage market. So the SDD market, at $.30 per GB, would be $60 billion totally converted. With the best of today's NAND technology the SSD market would be $120 billion....that's too much. $60 billion, filled with advantages, is in the right ball park.
SSD would fill the excess Intel capacity, but maybe they have some other idea for that capacity. We will find out.
Intel is a buy and 2014 is the year.