In addition to experimenting with different game development models, everyone is also continually experimenting with ways to keep development costs down.
As noted, art resources are jumping fast. It takes a lot of effort to detail a space to look good in current resolutions. Yay the hardware can push enough polys and texture density to achieve that level of detail, but painting it all in takes enormous effort.
And memory! The effort we all go through to crunch down the resources enough to fit in memory, to not take forever reading off disc, to not be noticeable as lazy-loaded resources “stream” in…
As the world gets more detailed, everything within it must improve in fidelity to match, lest it stick out as inappropriate. So before where an object might fall at a simple linear velocity, bounce once and stop, now we have licensed rigid body physics solutions that can simulate what an object roughly bounded by jointed capsules would fall like as it tumbled off the railing to the stairs below. Except that thing is the ragdoll of a humanoid, and the joints can’t just bend without constraint and still look natural, so now you need to feed constraints back into the solution…
And why didn’t that little wooden chair shatter when I hit it with the shotgun? Maybe we should look into that materials solution the Star Wars guys used (Force Unleashed). It’s pretty expensive; should we try to roll our own?
And so on.
Art outsourcing is growing fast. For FMVs, the results are pretty good. Blur is a good example. In-game outsource assets generally come back with more mixed results. The Gamasutra post-mortem articles frequently bemoan the communication and quality problems arising here and swear they need to do it better next time.
On the engine side of things, middleware is equally so-so as a cost-cutting measure. Witness all the times Epic has had to stress that licensing Unreal won’t save you any development time. The costs of making an engine with the expected levels of visual fidelity are high and getting higher.
I’ve seen 2 intriguing developments recently that give me hope for some reduced costs. First is the growth in high quality procedural generation. That technology looks like it’s getting to the point of feasibility again. Second is smarter use of template assets, as seen in Infamous and Borderlands. Stamping out bunches of prefabs with a modest toolbox – building worlds lego-style – will save big bux when applied smartly.