I'm not super familiar with the numbers, but I believe the "billions to develop" number is a development number, if a little juiced. I think they take the total amount of R&D spend and divide it by the number of drug launches for a given period of time. So it's a good idea of how much money you have to spend to get a new launch, but it is higher than the cost of a single project, the majority of which are terminated before launch (many well before clinical trials).

After you spend a couple billion per launch to make a new drug, then you spend another couple billion on marketing. Again, I could be mistaken, but I'm pretty sure that's the position of pharma.