Tom Foremski, Silicon Valley Watcher

On Monday I met with Intel's rising star senior vp Anand Chandrasekher, head of worldwide sales and marketing. Mr Chandrasekher has been in the job less than a year, previously, he headed Intel's Centrino group, one of the most successful Intel businesses of all time.

The Centrino mobile chipset business has been a huge cash cow for Intel. It caught the shift from desktop to notebook just at the right time. And Centrino has been instrumental in pushing Intel's gross margins back up into the 62 per cent range.

These are software industry type margins yet they come from a chipmaker which has to juggle billions of dollars in capital costs in predicting future market needs. On top of that, it has to operate the most expensive and most complex industrial factories of our modern age: the chip fab.

In fact, this is the secret to Intel's success. It is not that it is a microprocessor maker (that's just the app) it knows CMOS like nobody else. CMOS is the dominant process for making advanced, highly complex, microprocessors and other types of chips.

And Intel is utilizing nearly every element in in the periodic table to create new materials that enable it to double the complexity of chips every 18 to 24 months.

Here are some of the highlights from the interview:

[Mr Chandrasekher is Indian, and during our meeting, one of his aides kept him fed with cricket scores, there was a big game on :-)]

Low-power high-performance

Low electric power consuming chips are a prime focus for Intel, says Mr Chandrasekher. "We did a study about five years ago and found that our customers were already complaining about high electric power costs in their IT data centers, so now, our designs are totally focused on reducing electric power usage and performance per watt of power."

Intel joins Sun Microsystems, Advanced Micro Devices, IBM, HP and smaller competitors such as Transmeta, in the hot market for low-power computing systems. This is especially important since utility computing will likely become the most important IT architecture of the second half of this decade. And this requires data centers with tens-of-thousands of servers.

Mr Chandrasekher says the future of low-power chips lies in multi-core processor designs, getting rid of older design features such as deep pipelines, plus other tricks, and power consumption falls with each tweak.

What interested me the most, was whether Intel would be able to walk away from its strong belief that you have to have a big, powerful general purpose microprocessor on the PC or notebook.

We know that very soon we will have ubiquitous broadband access--wired or wireless--nearly everywhere. [Yes, Silicon Valley will be the last to get ubiquitous broadband, ironically.] That means we can offload general purpose application processing to central servers and use local computer devices to render the user interface and content screens.

Web services and technologies such as AJAX, enable software applications to carry their own processing within their user interface. You can provide users with a strong PC experience using just some of the "basic" PC chipsalong with powerful inexpensive graphics and audio chips..

All that's needed, imho, is a simple PC/notebook type device that is great at rendering and displaying graphics, video and text, and interactive elements--all within a familiar browser interface.

Why pay a couple of hundred dollars for Intel's latest microprocessor designs when all you really need locally is something like an ATI or Nvidia $50 multimedia/graphics chips with some minor bells and whistles. There is far less need for a general purpose, powerful microprocessor when you have easy access to big Iron over broadband.

A balanced client--neither fat nor thin

Mr Chandrasekher says that thin-client or fat server software models are in decline and that things will settle somewhere in between. And this is why:

"When computing was expensive, as in the 1960s, you had to get by with time-share computing. When bandwidth was cheap, it made sense to push the processing power to the clients and edges of the network. When both computing and bandwidth are nearly identical (zero) cost, I think client and server size will stay in the middle," say Mr Chandrasekher.

The $100 notebook computer

I take his point, but I don't buy it. I think you will be able to produce the fabled $100 notebook computer much sooner rather than later.

Mr Chandrasekher doesn't see a heck of a lot of value in a $100 notebook computer, ostensibly designed for the developing world.

"People with a $100 notebook computer will get the computer they deserve," he smiles.

He is convinced that middle-class families in developing regions of the world will pay much more for a computer because they see it as an important investment for their family and education of their kids.

I'm not persuaded. Time and time again, customers of the client-side device favor smaller prices. That's why the digital cell phone is the primary means of accessing the internet in much of Asia.

Mr Chandrasekher also talked about home entertainment markets, targeted with the VIIV architecture. Intel has ambitions to dominate this market--but so do about a dozen other large companies--approaching this market in their own ways.

Defending the margins

Intel has high gross margins to protect and so this makes it difficult to maintain high profitability in home entertainment systems where the margins are traditionally very cut-throat.

That will be the test of Intel's CMOS chip production prowess. Can it maintain the margins while at the same time churning out millions of VIIV based systems. This is a market that can be served well with specialized DSP and other chips--potentially costing much less than an Intel microprocessor.

That is, unless Intel can h cram ever more chips on a single die and come closer to a PC-plus-entertainment-technologies system that we could try out.