Why build an engine from scratch when you can just license it, learn it, and make the games? It's proven technology. There's wide support of it. No need to reinvent the wheel (again). Much simpler that way. Plus, with access to the source code, the dev teams can make whatever changes they need to make the engine work for them.

Also, by using an existing engine, there's less training that needs to be done for new employees. Chances are pretty good that new team members will have already worked with an engine like Unreal or Renderware or what-have-you, so they already know the basics.

The bottom line is developers need an engine that allows them to get their games done. "Innovation" doesn't necessarily mean a shippable product. Crytec and Far Cry aren't the norm. One game out of hundreds only means it can be done, but it is far from a sure bet.

Developing an engine *and* developing a game concurrently takes a very long time. You have to split your time between tool development and content development, and you have to somehow pay for that time. Most games take at least 1 to 2 years to produce. Add an engine onto that and development time can inflate to 2-4 years. The cost of something like that can run into the several millions of dollars (gotta pay the bills, pay your employees, pay for hardwares and software). Not exactly a cheap or easy thing to do.

Licensing engine might seem to cost a lot of money up front, but it can be worthwhile invest of time and money. Spend $100k on an engine that works right now and start making a game. Or spend a few hundred thousand dollars a year for the next 2-3 years to develop an engine that may or may not work all the while trying to make a game with an engine that isn't done. Do the math.

Man, after 7 years in this industry, I'm really starting to sound like the grumpy old guy. It's Monday. :p