The major point has been made: the hardware/ software demands of pulling 4K footage of a large sensor are pretty high.
That said, there are a couple other angles:
-- Canon clearly has the technology to do 4K-- e.g. the 1D-C. A high-end stills camera such as the 5D Mark III or the 1D-X could have been bestowed with the same powers (or at least smarter compression and downscaling of the 1080p image)-- but Canon chose not to. Why? Well, from their perspective, the real question is "Why should we?" When the 1D-C came out, it was the cheapest option for real, large-sensor 4K. It was also absolutely unique in its form factor and its ability to record 4K to CF cards. Yes, the Scarlet's out there too, but getting that camera in workable shape isn't cheap. It has some features that outclass the 1D-C (frame rates, RAW workflow), but unless you're making money off of these additional capabilities, the 1D-C offered the lowest TCO when it became available. 4K, from Canon's perspective, is something that enthusiasts want but that only professionals really need. As Sony and potentially Nikon and Panasonic get more aggressive in the stills/motion hybrid space, and as competitors such as Black Magic enter the cinema camera space, Canon might be persuaded to change its strategy. But Canon often looks at market transitions and waits until the last minute to make the change. Sometimes they wait too long (e.g. EOS-M) but their strategy is clearly to wait until a disruption is happening at a large scale, not to usher in the disruption themselves. Canon might still hold off on 4K below $10,000 for a while, given how obstinately the company has resisted 60 fps at 1080p-- but I can see where Canon is coming from when it basically says, "You want 4K? Are you a working media professional? No? Then what's the rush?"
2) The "consumers don't have 4K TVs" argument is a little dodgier. On the one hand, yes, it's true, if no one has the equipment to watch 4K content, then it's silly for Canon users (excluding certain professionals) to clamor for the feature. On the other hand, 4K TVs have become semi-affordable (e.g. you merely need to be well-off now, whereas you would have needed to be a top-1% earner to afford one back in January). Computer monitors and laptop/tablet screens are also pushing the resolution limit-- the 2.5K-ish Retina-level monitors are getting more ubiquitous, and all the chipmakers point out that their next-gen processors can handle 4K on multiple monitors. A lot of these computer resolution upgrades have to do with specialized industries--e.g. a stock trader who uses multiple monitors during work. At sub-50 inch screen sizes, the difference between 4K and 1080p can be hard to detect. Nevertheless, as 4K displays become commoditized over the next few years, the "you're a consumer and don't really need 4K" argument (which Canon has made) could become tougher to sustain.
3) Does anyone remember the alleged "30 fps burst mode" in a recent set of rumored 7D Mark II specs? I wonder if that's something similar to what the Nikon mirrorless cameras can do, and that the Panasonic m43 cameras do at reduced resolution. If the rumor is correct, I assume that AF would be disabled while this is going on, and that some sort of global shutter technology is being employed, which could have a few usability implications. Moreover, you certainly couldn't shoot a 4K feature in one-second increments. But still, if the rumor is correct, it would mean RAW 4K+ video, which, for a camera aimed at sports shooters, is pretty cool. It would also suggest nice things about Digic 6, and what it might be capable of. If the camera can manage 18-24MP bursts at high frame rate, then surely it could manage sustained 4K shooting for longer periods. The technical limitations would no longer be an excuse, so Canon's decision to include or withhold advanced video functions would come down to market positioning.
So, long story short, Canon isn't implementing true HD (let alone 4K) because it doesn't think it needs to yet. The technical challenges are undeniable yet nonetheless conquerable, so Canon's decision has been to milk the new tech in high-margin products, rather than to implement it at scale. It'll change that attitude when it thinks it needs to- and then the question will be, did Canon judge the market's dynamics correctly?