Now I like theory as much as the next (former) engineer, and I love to tell a good yarn, but as a statistician I prefer to test the theory on the numbers. To answer the question definitively, I aggregated the financial results of every U.S. and European semiconductor vendor over the time span of first-quarter 1998 to first-quarter 2004.

In theory, there are three major differences between fabless companies and IDMs: gross margin, R&D expenditure and the opportunity cost of capital. Since fabless companies need to pay a third-party foundry a markup for manufacturing services, theory holds that IDMs should have superior gross margins. But IDMs need to devote R&D to both process and product development, whereas fabless companies need only spend on product development. Finally, since IDMs need to borrow large sums of money from stockholders and bondholders, fabless companies gain an advantage on the opportunity cost of capital.

Since the advantages do not all favor one type of vendor, it is understandable that the fab vs. fabless debate rages on.

So, what does the data reveal? First, it turns out that fabless companies actually have superior average gross margins (47.3 percent, vs. 44.5 percent for IDMs), because the higher differentiation of chips from fabless vendors more than makes up for higher costs.

Second, fabless companies actually have significantly higher R&D expenditure (23.7 percent of sales) than IDMs (15.2 percent of sales). But while fabless companies pay an R&D price for product differentiation, they get a positive return on that investment.

Finally, using weighted-average cost of capital calculations, IDMs indeed pay a significant opportunity cost for owning fabs (10.1 percent of sales, vs. 2.4 percent of sales for fabless companies).

For the time period studied, fabless companies were marginally more profitable than IDMs.

Does this mean that IDMs should sell off all their fabs? No. The comparison is somewhat unfair because the products fabless companies are selling are better differentiated.

In the future, increased direct product competition will lower fabless gross margins and increase IDM R&D spending. As this trend hits both types of vendors' bottom lines, I would expect these business models to remain on par for at least the next five years.

Jeremey Donovan (jeremey.donovan@gartner.com) is chief analyst at Gartner Dataquest.