Why Intel's chipset shortage happened

WE ALL KNOW that Intel is facing a chipset shortage, the rumours are real, and the firm has confirmed it.

The underlying problem is twofold - capacity and attacking third party manufacturers, both of which were pretty stupid for Intel to do, so it is a completely self-inflicted wound. The situation is improving, however. Q2 will not be as painful, and Q3 will see 'they of the not dropped E anymore' back in almost full swing.

It is old news that Intel went after their partners in the chipset world in the name of profits, and they were quite effective. Serverworks was the best of the best a while ago, and then the guns were levelled at them after the Broadcom buyout, and they are more or less gone. SIS, ULI and Via are much smaller than they were, mostly grabbing scraps that Intel threw overboard. If it wasn't for AMD, these companies would be in serious trouble. Nvidia and Intel are smiling at each other through gritted teeth, and ATI sells anything in that market because Intel wants it to. It is all very ugly.

This situation arose because Intel wanted the market and the money associated with it, a short term gain at long term expense. It play some really heavy power games here, more or less forcing companies to buy a chipset with each chip because bundles were available, singles were not. Want CPUs in a timely fashion? Guess what you have to buy? This lead to a glut of gray market chipsets at blow out prices. Intel got the full price for the chipsets, with, ummm, "generously spread marketing dollars" to soothe the pain. See the Centrino programme for more.

Basically what happened was it killed its so called friends, and for a while, that worked just fine. It left just enough wiggle room for the Taiwanese chipset makers to survive, but barely, and I mean barely. Intel made sure there was no money in chipsets unless it decreed that there would be, and since it can shut that pipe down on a whim in less time than it takes to get a chipset to market, the sane vendors ran for the door. No one else was dumb enough to enter the Intel chipset market, and this situation is likely to continue. They planned the killing, and, er, executed.

Now, moving on to the capacity problem. A while ago, I commented that Intel is really really really good at planning, which is why I didn't think the shortage was the reason for the profit miss. I was wrong, Intel botched this one big time.

The official spin is that demand was so great that it was caught flat footed, and ran out of capacity. It is demand stupid, not any planning miscue. Looking at the numbers, this explanation holds up fairly well, Intel had sequentially greater sales last quarter, but not up to the wild and curious Wall Street expectations. So, it sold more, but not enough more, and not as much as it planned.

The problem is that these capacity planners have a pretty simple job - if you make one chip, you need one chipset. This is not advanced linear algebra here and a 1:1 relationship is pretty easy to plot out. If you have capacity for one million chips, you pretty well better plan for one million chipsets. Intel did not, and it cost them, it is a screw up, plain and simple.

Why not just change the mixture a little, make a few less CPUs and a few more chipsets? Intel makes chipsets on a -2 or a -1 process, so if 90nm is the cutting edge, the chipsets will lag by one or two process nodes, in this case on 130nm or 180nm. The current shortage of chipsets are all 130nm based, and now that Intel moved to 65nm for CPUs, it has room at 90nm to start making chipsets on that process. But the problem is that is not done with these chipsets yet, and it will be mid year before they start trickling out.

There is no more room at 130nm to crank up the volume, and a -2 process isn't worth spending lots of money to beef up capacity, especially since it is about to be yestertech in two quarters. Add that it would take longer to do than it would to finish up the 90nm parts, and you have a corporate situation best described as 'suck it'. There is no plan B.

There was a Plan B, and for years, if not decades, it worked wonderfully, it was called third party chipsets, from the like of ULI, SIS, and Via. They did well, and were the bread and butter of the computer industry for years and years. Intel got the steak, they got the side dishes, and all were happy. Then Intel got greedy, and squished them like bugs. When the time came that Intel had their proverbial nuts in a vice, there was nothing left to lean on. Even if there was, Intel had built up layers of deep animosity.

When Intel walked in and said: "We are in a bind, can you drop everything and help us?", the 'revenge' lights went on, and work commenced without any undue haste. Lunches became long lunches, and progress progressed, much as ice melts on a 33° F day. In the chipset makers defence, they could either help Intel regain marketshare, or help AMD make more. AMD is a partner here, and they do well together. The icing on the cake is that Intel knows it is temporary, and by Q3, the gravy train will have left the station heading back to Intelville.

So, aid rushed to Intel's side at a pace that make glaciers look like F1 cars, and they have two companies and a concept to blame. The companies are Intel and Intel, the concept, greed. Short term planning in an industry that requires long term planning. When you trip, and the ground is rushing up at your face, it is nice to have friends there to catch you. µ