There are those who have been denigrating marketing campaigns setting off with the goal of gaining "likes" on Facebook. And, to be sure, a million likes, by themselves, is going to do more good for your karma than your wallet. But many of the commenters come from a marketing background shorn of media buying experience, and those initiated into the dark art of the insertion order will recognize that likes create a multiplying effect for some media purchases.
There are two categories of Facebook advertising that benefit from likes: campaigns that involve creative targeted to social links and the links to those links; and campaigns that employ low cost-per-click bids that are traditionally starved of inventory.
In the first instance, some creative messages appeal specifically to those who have self-reported a relationship with the brand. Increasing likes increases this audience in a direct relationship. If a company has a marketing message that is more than twice as effective among those so linked, it may pay to acquire more linked individuals and market to them in that fashion.
In the second instance, Facebook holds opportunities for vast exposure very cheaply, so long as its system decides to reward particular campaigns with adequate inventory. The more you bid in CPM, the more likely you are to get more inventory. Although if the campaign is a social one, as most on Facebook are, the quantity of likes in the pool will have a large influence on how Facebook's system predicts the audience will react. With more likes, it is willing to throw more inventory at lower cost-per-click prices at a campaign in the expectation that the social links will cause more clickthrough, and more gross revenue.
Of course, the likes have to be "good likes," as opposed to the rash of fake likes that have been spewing weird good will over the social network over the past year.
The general quality of likes is becoming an important issue now with Facebook's much touted Graph Search, which depends on likes indicating a real relationship interest, which can be demonstrated to be untrue in some significant examples.