Companies have a lot of internal debate over this. EA has often touted their efficiency in producing a base IP that is ported across plaforms so that as long as the weakest platform sells enough to be even minimally profitable it lowers the overall cost of that base IP.
For example, say the shared elements of a game cost $3 million to develop. If that game is only on PS2 and Xbox, both of those version have $1.5 million charged against them in addition to their platform specific costs before they can show a profit. If a GameCube version is also done the shared cost per platform is only $1 million. That starts each version $500,000 closer to achieving the goal of profits. The accountants are all smiles.
The downside is the media production and PR costs for that weakest platform. If it doesn't at least break even it drags down the title as a multiplatform whole. So if the weakest platform's sales performance are in doubt and the product is expected to be a major hit on the other platforms, it may be time to give #3 a miss. Making this decision can be a lot easier if the project in question is a franchise sequel, as is the case here.
It says something I think, that my copy of Burnout is on GameCube. Not because of a platform preference but because that version turned up cheap while the others were still holding a higher price. Not a good sign.