Intel appears on the verge of achieving CEO Paul Otellini's long-held dream of diversification, moving beyond near-total dependence on x86 PC and server processors.
I wouldn't say they're there yet -- that's why I wrote "inches" in the headline -- but Intel has made significant progress in its bid to get off its traditional-processor addiction. The effort revolves around Atom, the downsized, low-power chip which has become the de facto device for netbooks. Intel has expanded Atom into a System-on-Chip (SoC) offering -- an explanation about that later -- enabling it to play in the burgeoning embedded computing arena.
First, some history. The idea that Otellini wants to broaden Intel's horizon's is nothing new. As I wrote when he became chief executive in May, 2005:
"A future-directed focus will likely become a hallmark of the Otellini regime, as he formulates a strategy for building up some of Intel's non-processor businesses--particularly, communications--that haven't been achieved their full potential."
That strategy hit a rut in 2006, when Intel sold its ARM-based XScale processor operations to Marvell Technology Group Ltd. for $600 million. The XScale devices were communications processors, which appeared in a variety of mobile phones, including some Blackberrys. At the time, I wondered why Intel was bailing on an effort which had previously been touted as a major communications push. In retrospect, I can see that the operation must have been a minor revenue contributor (Assume that the $600-million sales price was a multiple of revenue, and note that Intel's 2006 revenue was $35.4 billion.)
More importantly, Intel must've figured it would have required significant investment to maintain XScale's position in the rapidly evolving communications sector-- an arena where the product life cycle is faster than in PC processors, and, more to the point, where the timing of revisions is not dictated by Intel.
So Intel bailed. But now it's back, in a non-PC arena where it's nicely positioned to win. That area is embedded.
What's Embedded Computing?
Back in the day, the term "embedded computing" conjured up industrial, board-level computers usually controlling a manufacturing process in a factory. Adding to the esoteric patina, the boards were programmed by developers expert in C++, or, better yet, real-time assembly language.
Embedded computers are so-called because they don't look like general-purpose desktop computers -- they're typically mounted inside a box splattered with the logo of a manufacturer whose name is unfamiliar to the PC faithful -- hence the term "embedded."
In truth, embedded computing long ago shed its sole link with the factory. For years, there have been processors inside televisions and appliances. The difference today is that embedded computing on the consumer front is poised for a major expansion.
That's because consumer embedded apps like TVs, washing machines, and coffee makers are doing what desktop PCs did back in the 1990s -- they're transitioning from standalone bricks into Web-connected appliances. This is pretty simple technically but also profound in terms of market impact. We're talking about millions of televisions, cars, etc., as new-found platforms in which to stick a computing device.
Enter Intel, which intends for that device to be an SoC version of its Atom processor. "We have a vision of 15-billion connected devices," said Ton Steenman, general manager of Intel's embedded and communications group. "This is one of the big growth pillars of the company."
Steenman notes that Intel's already healthy embedded business is about to explode. "Between 2002 and 2008, we grew our embedded business between 16% and 18%," he said, noting that those percentages reflect a compound annual growth rate (CAGR).
Steenman won't provide hard numbers, but he said Intel is currently doing well over $1 billion in yearly embedded revenues. "Going forward, over the next five years, we predict we will grow 25% annually," he said.
In its public messaging, Intel is most enthusiastic about Smart TV. That's the Internet-enabled television platform co-developed with Google, Sony, and Logitech. Personally, I see Smart TV as approaching the opportunity from the wrong end. Consumers are OK with televisions offering a few extras, like Netflix access and maybe YouTube. Not so much with big screens grafted atop computers.
That was essentially the problem with Logitech's Google TV, which fizzled upon its January 2011 launch at the Consumer Electronics Show. A big impediment was a user interface only those C++ embedded programmers could love. (BTW, Google TV is essentially Smart TV under the Google brand, and it's not a stretch to foresee that term replacing Intel's desired moniker.)
Even if Smart TVs don't displace 40-inch Samsungs, it shouldn't put a crimp in Intel's embedded plans. Steenman said that Intel has "4,000 design engagements" -- people considering using Atom in embedded -- right now. Speaking to the potential success of Otellini's broadening-beyond-the-PC strategy, some 60% of those look-sees are from customers who have never bought from Intel before.
Inside Atom SoC
Earlier, I spoke about SoC. That's a fancy term in the semiconductor world, used to refer to silicon that's customizable in a cost-effective manner. Specifically, a bunch of off-the-shelf blocks of intellectual property (IP) are available and can be plopped onto a silicon die, according to an individual customer's demands. IP blocks refer to functional sections like a CPU, input/output handler, digital signal processor, network interface, etc.
It appears to me that Intel has slightly misappropriated the SoC term in applying it to what they're calling the "Intel SoC processor for cars, Internet phones and smart grid devices" [pdf press release].
These SoC Atoms are really regular, off-the-shelf CPUs. If they were true SoCs, they'd be different for each customer. However, they'd also be prohibitively expensive for most buyers. What Intel has done is to select the most popular functional blocks to create a "custom" part usable by nearly everyone. This is analogous to what Detroit used to do when they stuck a "custom" badge on a mass-produced variant of a popular sedan. The biggie in the Atom is an interconnect intended for easy pairing with a variety of input/output devices, the better to be useful in embedded apps.
No matter. Intel has brilliantly broadened its Atom brand beyond the netbook. That's something I never thought Intel would be able to pull off.
When Atom first appeared in 2008, I wondered how Intel was making any money on such a low-priced processor. The cheapest Atom netbook chip sells for around $29. The newer Atom SoC parts range from $19 up to $85, though volume prices paid by vendors buying in bulk are less.
In early 2009, Intel VP Stephen Smith told me that the company was achieving very high yields and that's how it was able to make a profit.
"We get something like 2,500 die locations on the wafer. We already have very good yields, very low defect densities," he said at the time. "We're not concerned about the cost of the Atom processor."
That's good news for Steenman, who in December was named general manager of Intel's embedded and communications group. He's got two potent levers to drive the embedded business forward: Atom's apparent profitably, and the huge expected uptick in volume, sparked by Atom's appearance in every greater numbers of connected consumer appliances.
"You'll have 15 to 20 connected devices in your home, easily," Steenman said. Intel's embedded business already does well over $1 billion in annual revenues. All of which means that Intel CEO Otellini's dream of diversification appears well on its way to becoming a reality.
Follow me on Twitter: (@awolfe58.)
What's your take? Let me know, by leaving a comment below.
Like this blog? Subscribe to its RSS feed: (here.)
My videos on ( YouTube.)
Alex Wolfe is editor-in-chief of InformationWeek.com.