Seeking Alpha
Profile| Send Message|
( followers)  

A decade of disappointment

Anyone who has watched Intel (NASDAQ:INTC) stock knows that it has been stuck hovering around $22 for the last decade. Sure, we've gotten paid a nice dividend for sitting around and we could trade the fluctuations to make a little extra money, but for those of us who are genuine fans of the company, it has been a disappointing time. Intel has had many quarters and years of top line and bottom line growth throughout the last decade with little increase in stock price, and many of the company's supporters have come to feel that Wall Street just doesn't understand Intel.

(click to enlarge)Intel Stock Price 2003 to 2013

Intel's many critics (Auguste Gus Richard of Piper Jaffray, Infoworld's Bill Snyder, FBR Capital's Craig Berger, and technology columnist Bob Cringley along with many others) have pointed to a number of factors, many of them off the mark, that have led to Intel's stock price stagnation. Some of the criticisms most frequently mentioned are:

Increasing capital expenditures in order to increase capacity and move to 450mm wafers at a time of supposed shrinking demand.

Sticking to Intel Architecture and not embracing ARM and becoming another open foundry.

Missing the "mobile revolution"

Reducing consumer expectations and margins by supporting "inferior" NetBooks and driving down prices of notebook computers in the process.

Not prioritizing energy efficiency and battery life in its designs (this is, in my opinion, the only real misstep made by Intel in the recent past)

Happily, I think our disappointment is nearing an end as Intel's investment in five key areas of research are finally starting to pay off and Intel is prepared for far more than just the next few product cycles; it is ready for a whole new class of computing devices.

Five parts of a revolution

In addition to their very successful and ambitious Tick-Tock innovation strategy focusing alternately on processor architecture advances and fabrication advances, Intel has invested heavily in five key areas that, I believe, will put it way ahead of most of its competitors within the next few product cycles and dramatically change the mobile computing device landscape over the next few years. Here are the five areas:

  1. Perceptual (or sensor-based) computing
  2. Machine Learning
  3. Intrinsic Security
  4. Ubiquitous Networking
  5. Code optimization for Parallel Processing

Let's look at each of these and see what they portend for the future of Intel, new computing devices and the stock price.

Perceptual (sensor-based) Computing: Anyone who saw the Intel booth at the recent Consumer Electronics Show or has read any of the papers/reports coming out of Intel's perceptual computing research saw the very beginning of what is coming: natural interfaces well beyond simple touch including complex gesture and voice recognition with inflection and emotion detection, but this barely scratches the surface of what is being invested in and what is coming.

If you take a look at the average, current generation, smartphone you will find upwards of 15 sensors; some of the sensors have their own silicon pre-processing the information they collect while others feed, more or less, directly into the main processor (through North/South Bridge silicon). For example, many smartphones contain two cameras, a microphone, a gyroscope, accelerometer, touch screen, light/contrast sensor, GPS, 3g/4g cell signal transceiver, near-field communications, Wi-Fi, blue tooth, a volume control, and an on/off button (yes, that is a sensor too).

Each of these sensors and accompanying digital signal processing silicon (DSPs) are often made by different companies (Analog Devices, Samsung, Infineon, TI, etc.) with different programming schemes in their firmware and different interface specifications. Now imagine instead of 15 or 20 sensors there are 150 to 200 sensors that read your eye position, head position, heart rate, breathing, eye dilation, oxygen levels, body position by communicating with micro-sensors embedded in your clothing, etc. Think about how much silicon is involved now. Can you see why Intel might think about ramping production of very small, very low power processors for the future while providing a unifying architecture (IA) and an application programming interface that millions of developers are already comfortable with into each of these semi-specialized processors?

So why would you want your smartphone to have access to all of this sensor information? The answer is everything from medical condition monitoring & emergency response services to reading your emotional reaction to a new outfit you are looking at in a store and recording that reaction into a personal preference database in order to help you pick out clothing in the future. Imagine the remarkable entertainment experiences such an integrated sensor system could provide (along with an equally impressive display/audio system).

If Intel were to produce even ten percent of the silicon needed for this sensor-heavy computing model then that would be a nine-fold increase of their total addressable market today (which they have not come close to exhausting).

Machine Learning: Basically machine learning is all about recognizing patterns, storing those patterns, and acting intelligently on those patterns when the occur. In order to bring about robust natural interfaces, make intelligent suggestions to users, and provide complex, enriching and personalized services, computing devices need to integrate huge masses of fuzzy, seemingly inconsistent, and idiosyncratic data often in near real-time.

Essentially, advances in machine learning are needed to intelligently integrate all the information gathered from the sensor scheme mentioned along with stored information about a user from the vast array of accounts, preferences, and personal information stored in cloud based and local databases. Intel, adding to an already significant investment in the field, has just recently invested millions of dollars in an Israeli research facility to advance machine learning and is clearly dedicated to research in this field.

Intrinsic Security: Intel's $7+ billion acquisition of McAfee was not just about getting a software cash cow to diversify its revenue streams. It was about developing intrinsic security for computing devices. In a world where your computing device is going to have access to biomedical, financial, personal, and location information then a built-in, low error, fast and automatic security system is of paramount importance. You and your device will share the most intimate knowledge of you and your social, home and business life. The device and the services you use must insure that you are you, that your accounts are yours, that stored data is not easily accessible by others, that you have full control over your personal information, and that malware is prevented from gaining a foothold and not simply detected when it arrives. The combination of McAfee's security research put directly on Intel's silicon could make for an extraordinary security infrastructure that can be scaled from massive server farms to individual devices.

Ubiquitous Networking: While many technology companies (including Intel) are investing in big networking: things like cloud computing, web services, big data, and the like, Intel is also investing heavily in small localized micro-networking. Intel's researchers have been working on small RF powered networking/computing devices they call WISPs (Wireless Identification and Sensing Platform, sometimes also referred to as motes). These tiny devices can be embedded in nearly everything from inside clothing & grocery packaging all the way to under skin devices. They are similar to RFID tags but are active computing devices capable of limited computing and signal relaying. If these WISPs were distributed on product labels, in a store say, it would allow your mobile device to pulse an RF signal and read the prices and descriptions of products you were standing near allowing you to find that book you were looking for that's hidden behind the other books you can see. Your device could read the device of the attractive person you are standing behind on the coffee line (without a cell or wi-fi signal) to see whether they are single or married (if they so choose to broadcast this information). You could also get internet-like networking from very remote locations by sprinkling these WISPs over the area and signals could propagate from one WISP to another in a tiny, micro-network to the nearest internet backbone or cell tower.

Code optimization for parallel processing: As a sometimes systems software developer I have had several opportunities to experience Intel's progress on optimizing software for multi-core, multi-processor, and massively parallel systems. While Intel continues to push developers to learn parallel and asynchronous programming techniques and encourages such programs of study in multiple universities, it is the automatic conversion of serial code to optimized parallel executable that is one of the true Holy Grails of modern computer science, and it is here that Intel's experience with high performance computing really shows. Over the years Intel's Parallel Studio, among other impressive tools, has gone from making fairly obvious suggestions on how to change code to better make use of extra processors and hyper-threading to now, showing the interested developer how to parallelize things he/she didn't even think was possible.

Integrating information from hundreds of sensors, accounts, networks, and devices in near real-time is going to take tremendous processing power and multiple parallel processing is the most sensible way to get that kind of power. While many of Intel's competitors have focused on optimizing data movement and media streaming, Intel has not shied away from high performance processors nor the research and tools to back them up.

Timelines, Predictions and Price Targets

So, while all this research is well and good, where does this leave us, the technology investor? Here is my prediction for a timeline of product introductions, new features, and price targets for Intel:

Current to 18 months out: Product cycles are now as fast as 4 to 8 months instead of 18 months so we are looking at, perhaps, three generations within 18 months. We are already seeing natural user interfaces making progress from the novelty phase to full business and personal use. Touch has matured, voice recognition and handwriting recognition have gone from the high 80%s in accuracy to the high 90%s and some are even crossing into the 99%+ range. We will see first major product wins for Intel's low power products. We will likely see early integration of data from multiple accounts like social networks, shopping sites, and location services leading to devices that offer predictive analysis of users' preferences that are actually accurate and useful to the user. Price target at 18 months: $30-35.

18 to 24 months out: I think we will see fruit from the McAfee merger with mobile, desktop, and server processors with robust security features; I believe that this is a prerequisite to the deeply individualized computing experiences I am predicting. I think we will see the beginning of limited rollouts of active RFID chips/WISPS and the 22nm processes will start to become standard. RF readers will start to be integrated into more expensive smartphones, tablets and other computing devices and diverse services that make use of the readers will emerge. I believe some of the less powerful/less specialized sensor interface chips & micro-controllers will start to be replaced by IA chips. This will dramatically increase the volume of processors sold, though the margins on these small processors will be somewhat less. Price target at 24 months: $ 40-45.

24 to 36 months out: Sensor proliferation will really get going; we should start to see up to 40+ sensors per device and machine learning software starting to sense and interpret physiological and emotional states in useful ways as well as biometrics finally reducing our password load in a real way. Margins will drop by about 20% from current levels, but volume will increase by up to 2.5 times current levels. Price target at 36 months: $45-50.

36 to 60 months out: Five years out is a long time in the technology world and I have little confidence in the time line at this point, but my guess is our more expensive devices will have non-local displays (eye pieces, image projectors in glasses, etc.) and non-contact interfaces (gesture control, eye tracking, voice, clothing based touch pads, etc.) fully integrated though not yet replacing touch screens, keyboards, or even mice. Intrinsic security will have matured sufficiently that people are comfortable having their devices "know" them and the era of having a mobile device be a true digital, personal assistant will finally have begun. By this time, if all goes well, Intel should be producing about 5x the volume it currently is in processors alone as well having successfully entered several consumer products markets. Price target at 60 months: $60-75.

Source: Intel Perfectly Positioned For Next Computing Revolution