My readers know that while I am completely on-board and bullish about Intel's (NASDAQ:INTC) 4th generation Core processors (codenamed "Haswell"), my belief is that these will merely put the company's PC/convertible business back on-track to mid single digit growth. Now, this is certainly a wonderful business, and given that Intel is a mega-cap with a $54B revenue base, this is certainly not small feat. However, the real show-stopper - and what I believe will put the client business on the double-digit CAGR map - will be products based on the "Silvermont" Atom core, and it starts with the tablet-oriented system-on-chip called "Bay Trail".
There's A Lot Of Volume Down Below...
Anybody who has paid attention to the investor presentations that Intel has given at countless roadshows will be familiar with the following slide:
The ideas here are simple:
- From $199 to $550, we will see "Bay Trail" powered tablets, notebooks, Chromebooks, and "2-for-1" convertibles.
- From $550 to $999+ we see premium "Haswell" powered devices
So, at the high end, we see that Intel will be selling its "Haswell" parts. The people who want/need performance, or simply want to have a higher end device, will still keep buying "Haswell" based systems. The chip dies are more expensive to make, and as a result, cost more - the customer pays for performance.
But my guess is that the new, sexy designs of Ultrabooks and even non-Ultrabook higher end notebooks will get back onto a low to mid single digit growth curve for the following reasons:
- The designs and form factors are just "sexy". While "Sandy Bridge" and "Ivy Bridge" notebooks went a good way to making notebooks thinner, notebooks still suffered from really lousy displays, lack of touch, and yes even bulkier designs, the "Haswell" based designs being demo'd at Computex are absolutely gorgeous
- Battery life gets a huge boost, thanks to a dramatic reduction in idle power and new low power states that keep the processor from wasting energy in periods of little operation
- Fanless "Core" based tablets are now possible, and were even demonstrated with Intel's reference "North Cape 2" design as well as a design from HP (NYSE:HPQ)
That being said, a massive volume spike isn't going to come from $550+ devices; Intel will need to play aggressively in the lower cost markets, too. Why is this?
My view of it goes like this: most people have a limited amount of money to spend, and at the $399 price point (where entry laptops and high end tablets coexist), people will want to buy the best product that they can. Today, a $399 Google (NASDAQ:GOOG) Nexus 10 is a much "better" device in many ways than the generic $349 laptops that you find on Newegg.com. Here is a set of reviews of a $349 laptop that highlights just why low end PCs are dying:
While I urge you to read these "reviews", the main points are:
- Lousy touch pad (user experience fail)
- Bad WiFi
- Feels cheap
From this we can develop the following broad congruence classes for what fundamentally needs to change in the PC industry:
- User input is a big deal and should be intuitive/bug-free
- Connectivity is integral to today's PC user experience
- Speed is still important (both graphics and CPU) - "good enough" is only so until the user can have something better
- Build quality is important
Is it any surprise that even at $499+, Apple (NASDAQ:AAPL) was able to sell a boatload of iPads, and then when the iPad mini offered the same quality at $329, those sold like hotcakes, too?
The key take-away is that Intel needs to get a high performance, cheap-to-manufacture chip for the low end to the market as soon as possible in order to free up both bill-of-materials and power budgets towards making the lower end Windows notebooks/convertibles as well as Android notebooks/convertibles that are still high quality machines. The problems with the current "Celeron" and "Pentium" low end products that are castrated "Core i3" parts are the following:
- The die sizes are larger than what the ARM guys have by virtue of the fact that while these are disabled/less functional parts based off of "big core" dies
- The power consumption - particularly at idle - of these castrated "Core" processors was high and cheap PCs don't have a lot of budget for a battery, which meant that low end PCs have really lousy battery life
- A little known fact by the investment community: Intel's "Celeron" and "Pentium" low end chips - just like their high end brethren - also require a separate chip known as a "southbridge" that controls USB, Serial ATA, and other peripheral functions. Not only is this another chip to worry about, it is usually built on a generation or two older process technology, so this itself usually consumed plenty of power
- The "southbridge" also wasn't cheap (~$30 - $40), so that ate more into the bill-of-materials, further compounding the problems that I mentioned above of crappy displays and even crappier battery life
Now, interestingly enough, it seems that Intel will be taking the "castrated Core" based Celeron and Pentium chips to the woodshed and will instead replace them with a single-chip, Atom-based processor known as "Bay Trail-M":
This, I believe, has the following effects for Intel:
- Intel will be able to command roughly the same, if not slightly more, per smaller-die SoC since Intel is now selling one - probably overall smaller - chip. This means that gross margins stay intact
- Intel's OEM partners will have bigger power/space/money budgets to make even the lower end designs more compelling and drive volume at the low end of the compute continuum
- Intel's SoC volumes will skyrocket, blended ASPs will decline (by the law of averages, not due to margin erosion at the high end)
- Intel will need to find something to do with older generation fabs that aren't building chipsets
I would like to elaborate further on this last point, as I believe that it has some serious puts and takes from the perspective of the investor.
The Older Fabs - What To Do With Them?
Remember when Intel said that its fabs were at ~50% idle a quarter or two ago and that Intel was taking down older generation capacity for 14nm? You don't think that it was taking down 22nm plants do you?
No way. The stuff that goes first is obviously the older generation - think 65nm, 45nm, and 32nm. Intel has been gearing up for the launch of its 22nm SoCs, server parts, and Haswell - so taking down 22nm capacity would be absolutely silly! The capacity that was probably on the chopping block was 32nm, since the capacity here would be used primarily for chipsets accompanying the 22nm server and "Haswell" parts. These chipset die are much, much smaller than the CPU die and as a result not as much capacity here is needed. Current 32nm Atom volumes are pretty negligible now and should trend to zero as all of the Atoms for phones/tablets/micro-servers shift to 22nm parts.
Anyway, so I believe that instead of building chipsets for Ultrabook-oriented "Core" and "Atom" processors going forward, Intel will use the fully depreciated n-1 generation fabs to flood the "cheap" tablet/smartphone markets with older generation, cheap-to-make, older generation parts to gain market share, particularly in emerging markets. This would make life very tough for the likes of MediaTek and AllWinner as Intel's older generation parts would provide better performance and be priced to move, while the high end commands all of the pricing power.
There is probably enough volume in entry smartphones and tablets to more than offset the loss of PC chipset volume, and I still think that traditional high end desktop and parts of the server market will still rely on a two-chip solution, although in a world where even the data-center requires massive power efficiency/TCO reduction, Intel will - in a bid to keep its low power, very efficient Atoms - give the Xeon line every competitive edge with respect to integration/power efficiency to avoid cannibalization.
Intel will be able to enable much higher quality and much lower price points for its entry level PCs/convertibles going forward, and I think that this not only will help Intel solidity market segment share across the computing continuum, but it will ultimately help Intel continue to scale the volumes that it needs in order to justify investments in future fabrication technology. When wafers grow to 450mm, and when transistors shrink massively in size, Intel's going to need to be selling a lot of chips, and the cheaper the devices that Intel gets into, the more volume it can ultimately drive.
Intel realizes this, and that's why "Bay Trail" was pulled into 2013 when it was initially going to be a 2014 event. This Christmas, everybody will be buying these cheap, high quality, touch enabled laptops/convertibles/all-in-one PCs, and it will be enabled thanks to "Bay Trail". "Haswell" keeps the high end of the computing market alive, particularly in tablets and ultrathin notebooks, but there is just so much more volume to be had at lower price points that I believe that investors cannot afford to ignore the Atom story.