Future Ala Carte = Content from internet = CUTTING OUT MIDDLE MAN (cable companies, satellite companies, FIOS, U-Verse, etc). This means cheaper pricing.
It's important to realize that IP delivery of broadcast video is still very early in its technology adoption lifecycle -- the pricing that is in place today isn't indicative of what a solution would cost when fully scaled. To parallel a similar scale-constrained techology, folks running biodiesel vehicles are getting by incredibly cheap right now because they're able to tap into the discarded cooking oil from fast food restaurants for little/no cost. If too many people convert to biodiesel, now restaurants starts to face more demand for their old cooking oil -- which creates a market for that resource and pushes up prices. You also have the scale issue that there isn't enough cooking oil to support more than a fractional percentage of vehicles making the conversion.
Content from the Internet doesn't necessarily mean cheaper pricing; the current distribution methods involve unicast IP delivery feeds. That means that my video feed is separate from your video feed, so the amount of bandwidth required grows linearly with every single viewer. Compare that to Cable/Satellite/OTA which have fixed costs for transmission regardless of how many people in the broadcast coverage area are viewing the feed.
Internet-based video delivery is about 4 legitimate and 1 questionable thing:
1) Time shifting (watch what I want, when I want it)
2) Place shifting (same as #1, but not primary viewing location)
3) Archived Content (missed an episode, go back and download it)
4) Niche content (Limited distribution)
Questionable) "Free" programming
Time shifting is by far the largest driver towards Internet-based delivery, and it also represents the most inefficient use of network resources. You're taking content that is already being digitally delivered to your house via OTA ATSC feeds, cable QAM feeds, and satellite QPSK/8PSK feeds.... and transferring yet another copy of the same content over the network. Time shifting is better accomplished through local capture and playback using an already established stream as the source.
Internet delivery has a promising role in all of the other options, but re-inventing the wheel for primary / time-shifted delivery would push costs way up, not down. I see networks banding together to produce a cheap/free ad-support DVR available to customers long before they go exclusively to the Internet for distribution.