Intel In Slide

Things should be rosy for Intel. It's decided that the future is multi-core, and the rest of the industry agrees. Microsoft is doing the right thing, up to a point, over multi-core licences while AMD is right up there singing the same song.

Yet this is scant comfort for a company that does not seem able to make new chips or sell the ones it has. Intel's 64-bit chips are expensive, late or absent, while its 32-bit parts have run out of steam -- it has canned the next speed increase -- and are at sea in an ocean of shrinking margins. Those dual-core chips? Late. Only the mobile parts are holding their own.

Intel has big problems. It has around a complete quarter's worth of inventory on the books, which it has to shift by cutting margins or writing off. Either way, that costs. Intel's operating costs are static and do not change much no matter how many chips it makes or does not make, so it can't scale back on production and feel the benefits -- at least, not unless it starts hacking at capacity by closing fabs down.

It could innovate out of the hole, but here too there is a sad litany of projects stubbornly refusing to come to fruition. It has new ideas for memory, embedded processors, display devices, wireless and photonics -- but it has few new products. The technology is demonstrated, great things are predicted, and silence descends. You cannot grow a market that way.

It could market what it has better. Again, the mobile side has done well -- proving that with sustained spend and a coherent message, it can create and sustain a brand. Or perhaps that's just because there is no competition. On the desktop and increasingly in servers, AMD has been there a bit earlier, a bit faster, and with a bit more panache. Even AMD's chip-naming scheme is smarter: you know that if you have a chip with 3400 in its name, you'll get something better if you go to 4000. Try unpicking Intel's three-digit scheme, seemingly designed to hide features.

If Intel cannot invent new chips, it is going to have to reinvent itself. That will be a painful, costly process, but necessary. It will have to listen to the market rather than dictating it -- it could start by moving the Pentium Ms to the desktop. We don't want faster chips. We don't want hyper-threading. We don't want dual-core. We want low-power, low-footprint desktops with no fans but plenty of multimedia grunt.

Then the company has to take a long, cold look at what the market will look like in two years time -- realistically, the shortest time in which it can rebuild itself and the furthest anyone can look ahead -- and what the best fit in terms of product and production capacity will be.

Hard decisions have to be made.