Early trips would of course not have a large crew, since they'd be mostly focused on setting up a base, and be more about cargo and equipment. Saying "several tens of millions" is because that's generally what it costs right now to get someone to low Earth orbit. If we can get early trips to Mars down to what it would cost for a current orbital mission, that would be very impressive.
Start with ~$30 million (based on "several tens of millions") to get a person to orbit using current tech. Current tech doesn't use reusable rockets, and the cost of building a rocket greatly exceeds the cost of the fuel for the launch, so we'll assume fuel costs are 0. If you make a rocket that you can reuse for 10 launches, your price tag is down to $3 million to get a person to orbit. The next step is to make a bigger rocket: currently the Soyuz rocket is the only rocket taking people to the ISS (assuming this is roughly what you mean by getting someone to low Earth orbit), and it can only take 3 people at a time[0]. If you make your reusable rocket big enough to take 100 people at a time, your price tag goes down ~33x to $100k to get a person to orbit. Of course, a bigger rocket will probably cost more to make than current "small" rockets -- but a big reusable rocket would have to cost 330x more to build than a modern small rocket for future seats to cost ~$30 million per ticket.
So even the first trip to Mars should be well under $1 million per person as long as they make the rockets big and reusable.
[0] http://www.nasa.gov/audience/forstudents/k-4/stories/nasa-kn...
Even if the cost for the first few trips was $50 million per person, there'd be no shortage at all of people who would be paid for by governments or organizations with that kind of money. Might as well use it, and apply these funds to expanding the program.
Full reusability is really the essential part of this system - you can't afford to throw away any stage of a rocket this big. Fortunately, the larger the vehicle, the easier reusability tends to be (up to a point of course).
