- Intel is very transparent about their future plans.
- Did TriGate transistors affect the Intel mobile effort?
- Expect a comprehensive mobile solution in the near future.
Intel (NASDAQ:INTC) has been the poster child for pundit abuse for a long time. The truth is that Intel has been dead money since the tech wreck of 2001. The stock was significantly overpriced back in those days, and many holders took major losses from the all-time high price levels of $70+.
An analysis today indicates that the stock has more than grown into its valuation and is currently cheap when considering some highly probable developments in the near future.
The biggest criticism of Intel is the perceived miss of the mobile business over the past two years.
It might be that this miss was an unavoidable consequence of the move to TriGate transistors in 2012.
So, let's start with a look at how well Intel is able to look into the future of their business. This link is to a white paper titled Platform 2015 written by six Intel technical fellows (technical equivalents to Vice Presidents) in 2005. It is a technology outlook on the computing industry for the ensuing 10 years. The white paper is no longer accessible on the Intel website…or anywhere else, for that matter, except for the University of Potsdam website. If you want to keep it for future reference, you might want to print it. No telling when it disappears altogether.
The predictions are uncanny in their accuracy:
- The very first macro trend the paper discusses is pervasive connectivity through cell phones. They knew that wireless notebooks, handhelds, and wearable systems would be in our future. So, Intel didn't miss the mobile business by accident or ignorance.
- This group recognized the coming change in computing form factors and price points driven by the continuing globalization of consumer computing. I'm sure that Intel was not "blind-sided" by tablet computers.
- They recognized the coming "software as a service" supported by a faster and more functional Internet, which, of course, needed higher performance server chips.
In 2005 these experts knew that the power levels of computing devices would have to come down dramatically. They saw "always on", the security threat via mobile devices, sensor nets and the need to include these functions on the computing chips themselves. Intel saw the expansion of chip level multi-processing (CMP), with dedicated cores running graphics, speech recognition and communications protocol processing (LTE?).
Through this paper, Intel told us where they were going. The memory bottle neck would have to be solved with Gb+ size near chip memory systems (Crystalwell), High-k/metal gate and tri-gate transistors are mentioned. Possible futures include III-V transistors, carbon nanotubes and silicon nanowires. They expected to build chips with billions of transistors on a square inch of silicon, and they have done both.
This is an amazingly predictive white paper.
So, why is Intel an also-ran in the mobile chip business?
The problem appears to be with TriGate transistors. Since TriGate transistors are fundamentally different from planar transistors, the analog functions required for a mixed mode mobile SoC, in many cases, don't exist. They need to be developed and that is very time consuming in terms of scarce engineering talent. In fact, that Intel was able to produce a 22nm TriGate version of Sandy Bridge for the Ivy Bridge "tick", seems now to be a Herculean task.
Intel had no choice but to convert to TriGate transistors in order to keep the scaling progress on x86 chips intact. After all, the x86 chips are the franchise and generate virtually all of Intel's profits. The problem is that mobile had to wait for the required analog functions to be designed, developed, and verified for producibility. This seems to be borne out since even the third generation Intel mobile APs, Sofia, etc. and LTE chips are being built on the TSMC 28nm planar process, where analog cells exist.
There is an even a bigger problem with analog and TriGate, and that is that for each TriGate shrink the analog design process appears to have to begin all over again. This can't happen in the Intel world. There is an alternative that is being developed by Intel that would instantly put them in a leadership position in mobile. That alternative is full digital radio. Nearly two years ago Intel demonstrated a full digital Wi-Fi transceiver. At that time there was some chatter about commercial product for mobile digital radio in the 2015 timeframe. Digital analog and digital radio will allow Intel to continue scaling x86 CPUs and mobile SoC to at least the 7nm node.
If the above is true, and I believe it is, the implications are huge. What I have just described is a conscious decision by Intel to temporarily pass on the mobile business in the interest of creating more functional CPU and server chips, where their profit is. I am supposing that this analog problem is also keeping TSMC and other foundries from making the conversion to their respective finfet processes. Certainly TSMC could move to finfet rather than enduring the pain of squeezing a 20nm node out of planar.
With the release of Broadwell, Intel will be a Tick-Tock cycle and a half in TriGate ahead of the next closest competitor, with about a three-year head start on solving the analog problem with TriGate analog designs or, more likely, digital, scalable solutions to the entire CPU-SoC problem at two to three nodes ahead of the competition.
OIA loss: the Other Intel Architecture group has lost an aggregate of $3.8 billion on $8.4 billion of revenue for 2012 and 2013. That almost doesn't seem possible unless a great deal of that loss is embedded in high R&D allocation for the mobile chip business and the digital solutions described above.
In conclusion, it appears that Intel knows, with a high degree of accuracy, where it is headed and the company's degree of successful execution is admirable.
From the white paper one might predict full-function Intel mobile application processors with embedded digital radio and on chip DRAM.
Upon the word of a successful programmable digital radio LTE solution Intel's stock price could easily double.
This is not a play without anxiety.
I play it with calls or LEAPS and have lost 100% for two years running. For those of you who know how I use options, you won't be surprised by this; I expect to only win one out of three option plays. The losses are often 100%, but the gains are often 10 to 50 times.