Seeking Alpha
Long/short equity, contrarian, independent research, tech
Profile| Send Message| ()  

I make it my priority to monitor (and when appropriate, write about) every little rumor, leak, and whisper when it comes to Intel's (INTC) product plans, as I believe that a significant portion of my readership - particularly in the investment community - relies on my columns for up-to-date information concerning their investment. Very recently, there has been a deluge of (seemingly) contradictory information with respect to Intel's 14nm "Broadwell" processor/SoC families, and I believe that I have been able to correctly piece the puzzle together.

What Is Broadwell?

Intel's processor development model is referred to as "tick-tock". The idea is that developing a new processor micro-architecture entails significant risk, and that moving to a new manufacturing process entails equal amounts of risk. Very clearly, moving to a new processor micro-architecture and to a new process technology scales risk quadratically, so what Intel does is that it essentially moves the existing processor design to the new process when it is available, and then in the following year, releases a ground-up micro-architecture based on the new process node. As far as I am aware, here is what Intel's current CPU core roadmap looks like:

(click to enlarge)

We just saw the release of the "Haswell" family of chips during 1H 2013 (although most systems will not really trickle out into the hands of consumers until 2H 2013), and accoridng to the most recent conference call, the company will be well underway in building the next generation "Broadwell" chip on the 14nm process. For completeness, here is the snippet from the call that verifies my claim,

We should get some good news associated with startup cost, and then the offset in Q4 is going to be, we should be well into the build of Broadwell.

Now, interestingly enough, while the CPU core in the "Broadwell" based chips will largely be a shrink of the "Haswell" design, the company plans to bring out an entirely new graphics architecture (called "Gen 8") with the new chips, which is why the "tick tock" model applies only to the CPU cores and not necessarily to the various system-on-chip designs built around the CPU cores.

Broadwell Delay Rumors?

On the most recent conference call, Intel CEO Brian Krzanich made the following statement about Broadwell's launch timeframe,

As far as our 14 nanometer Core launch in our - just our general product launch, I think what we've said so far is, first half of 2014 and we're not going to - we're not ready to give any specifics beyond that.

Of course, my friends over at VR-Zone leaked the following roadmap that shows Broadwell coming in the second half of 2014, not the first half as the newly appointed CEO stated just a week ago on the conference call,

(click to enlarge)

Now, interestingly enough, I actually would be willing to buy this particular rumor; that the Ultrabook-oriented Broadwell will actually be out in the second half of 2014, rather than in the first half. Why? It's simple. The "Haswell" for Ultrabooks, while likely in the hands of OEMs in the first half of 2013, was officially launched in Q3 2013,

(click to enlarge)

In order to ensure the proper ROI for the Haswell parts (which involves selling them for at least 1 year), Intel could not launch Broadwell for Ultrabooks in 1H 2014 given the "slip" (which I believe was intentional given that Ivy Bridge inventory needed to be cleared) in the Haswell for Ultrabooks. That being said, I do believe Apple (AAPL) will, as it did with Haswell, will ship Broadwell-based MacBook Airs in 1H 2014 ahead of the rest of the PC supply chain, since Apple is the largest customer for these low power, high performance parts, and is able to manage its inventory of systems well enough to be able to roll out a new device rather swiftly.

But, despite the "Apple" explanation, I still believe that Intel will be launching Broadwell en masse during 1H 2014 to non-Apple customers. Who, then?

Did Somebody Say Servers?

While the PC business is slowing down, forcing the lengthening of product cycles, the cloud/web 2.0 server business is booming. While Intel's highest end 2S/4S/8S monster chips take far longer to validate than their client cousins, so they are often a generation or so behind in micro-architecture and/or process, the more mainstream "micro-server" type workloads don't need such gigantic, difficult to make chips. Further, the growth, TAM, and looming threat of competition here is so large that the cost/benefits analysis probably shows that it is in Intel's best interests to push the latest chips out as fast as possible. It is my view that the very first Broadwell parts will be the Xeon E3 parts, followed closely by the Broadwell SoC and "Denverton" Atom parts:

(click to enlarge)

Further, it is important to note that not only is Intel selling the entirety of its LGA 1150 socketed Broadwell parts to the low end Xeon buyers, but it is even investing money to develop a special, highly integrated system-on-chip for the micro-server space based on the Broadwell core. Not only does Intel stave off the low end threat from the ARM (ARMH) players, but it also essentially locks in the market for higher performance, but similarly integrated parts.

Conclusion

So, there you have it. I believe that 14 nanometer parts will be available in the first half of 2014, but only the high growth markets where the process lead is essential to stave off competitive threats. AMD (AMD) is no longer a threat in the PC space, and the Atom parts for tablets and smartphones are quickly being accelerated to the leading edge (and it is my view that these parts will actually lead the higher end Core products soon). The leading edge process node makes the most sense in the highest growth, most highly competitive areas, where having performance and cost/transistor advantages is critical. It is clear that Intel is allocating its precious 14nm process to where it needs to go, and from a shareholder's perspective, this is precisely the right way to go.

Source: Intel: 14 Nanometer Mysteries Solved