As Jabil, Sanmina-SCI, Flextronics and others have noted, component shortages are becoming more prevalent as the recovery takes shape.

Worse, the pain is expected to increase as demand picks up. Some EMS companies are now using the dreaded “A” word — allocation. Others say they have been cut off by their distributors: orders placed weeks ago have simply been canceled.

For most of their history, semiconductor manufacturers made capacity expansion decisions outside the normal cyclical swings. In 2009, however, expansion effectively stopped. We will now pay the price — literally.

Memory, which has been in a glut for years, will suddenly become hard to come by. The December quarter was the best since the first quarter of 2007, as DRAM makers topped $8 billion in revenues. And as iSuppli noted last week, an explosion in smartphones is straining NAND-type ?ash. According to the research firm, each iPhone — the largest application for NAND — uses an average of 35.2 Gb of NAND. “This huge growth is likely to lead to some periods of undersupply for the year,” the firm says.

We could have seen this coming. This situation was set in motion following the drastic tech meltdown of 2001-02, when some companies were quoting material lead-times of up to six months! Since then, the supply chain has considerably tightened, and inventories have been kept low — too low, in fact, according to the Institute for Supply Management.

But not to me. I would rather see a constant capacity crunch than the wild swings of yesteryear. What companies must now resist is the inclination to take parts from wherever they can (thus increasing the risk of counterfeits) and the pull to bring on so much new capacity that they turn the tide and set in motion future oversupply conditions.