In brief, Theora is more open-ended (supporting up to terapixel resolutions!) but at the price of being far less optimized or well-supported for acquisition and delivery needs today. h.264, while processor-intensive and subject to long-term limitations delivers outstanding performance right now and is widely supported.

Also, Theora brands itself as ideal for web delivery with limited bandwidth leading to graceful rather than abrupt degradation, which is great, but of declining utility as bandwidth and reliability increase. It isn't pushed as a great choice for high-quality delivery, leading to a lack of awareness among content-providers who would rather use use h264 for everything from blu-ray to Youtube.

Now, that doesn't mean Theora has no future - as it matures it will become more desirable as a supported option, and the long-term architecture will help with that. Right now, however, it doesn't offer any significant technical benefit, like massively smaller compression or massively better image representation, compared to competitors. Making film and video is expensive and the patent licensing costs are such a tiny fraction of that cost that content providers and distributors really don't care. To them, Theora is a solution in search of a problem (indeed, many have never even heard of it). There is no economic incentive to deliver in this format, so they won't.

tl;dr programmers care about Theora, video people don't.