T’is the season for…semiconductor forecasts

T’is the season to be jolly…and to predict the next year’s semiconductor market.

KPMG does a regular survey of senior executives in semiconductor companies to get their outlook on the year ahead. The message this year is mixed. 41% of executives expected their business to grow by more than 5% next year, which sounds not too bad until you realize that last year it was 78%. The survey included over 150 executives, half from companies over $1B in revenue

The Semiconductor Business Confidence Index is at 46 compared to 60 the last couple of years (higher is better).

GSA also announced that so far this year there have been nearly 100 semiconductor acquisitions and mergers, down a little from a year ago.

Not surprisingly, the big drivers for semiconductor are wireless (probably over 2 billion cell-phones will be sold next year), computing (cloud server and laptops etc) and consumer (video game, TV, DVR etc).

The big indicator in semiconductor is capital equipment, because this is where companies put (or don’t) their money where their mouth is. Revealed preferences, an economist would say, which is much more important than what executives say in a survey. Here many executives plan to sit on their hands next year, with only a quarter expecting it to increase next year. Since a fab doesn’t get built in a month or two, this is an indicator of capacity that won’t be in place for 2013. There is also huge variability between different companies. Intel is clearly investing a huge amount whereas other companies are still switching to a fab-lite or fabless model where capital expenditure is not required (well, by the company: TSMC still needs to invest plenty). And the other big foundry, Global, is clearly struggling. They have a big fab in New York that is meant to come on line but they are also losing a lot of their semi-captive AMD business to TSMC.

The amounts involved are staggering. Intel is talking about $10B fabs. And because it can’t risk having just one fab (in case it burns down or something) they need to build two. But even Intel can’t fill two fabs hence all the speculation about Apple, Qualcomm, Xilinx and others using Intel as a foundry.

So semiconductor companies have had a two-year boom and are assuming next year will be slow. To some extent these can become self-fulfilling prophecies depending on how much capacity is put in place. If not enough, product will be on allocation even if there is end market demand, too much and prices crater creating end market demand.