You're missing about 90% of the factors needed, so it would take a long time to correct you. Looking at dividends is like looking at the previous two weeks of weather reports and concluding that it will be rain/sunshine/snow for the rest of eternity.

One thing most people are missing is the decline in share value that will happen in just over three years when the block reward halving happens. Because of AMs size, there's not much room for growth in percentage of network hash rate so the mining dividends will probably drop to about half. Because a lof of the current dividends is derived from hardware sales, which is a short-term opportunity, the real yield, taking into account the reduced hardware sales over time plus the reinvestment needed for hardware sale plus the halving effect, is far, far lower than you currently see.

In calculations I've done (and partially published) it's not very far from reality if you consider between 80-100% in yield of the percentage of the network that AM holds. In other words, if AM holds 100% of the network (which of course they never will), you can estimate a yield of 80-100% and if they hold 25% of the network, you can estimate a yield of 20-25% per year.

This figure depends a lot on what cost their next generation hardware will be (to them, not consumers) and how quickly competitors can delivery in scale. If for example BFL is able to deliver 65nm chips en masse before AM then you may consider reducing the above estimate to somewhere in the 65-85% area.

We don't know how much money AM has set aside for second generation chips, nor how much they will cost, but at some point, that cost will come out of dividends, so if they haven't set aside sufficient money for NRE, the dividends will need to pay for that in the upcoming months.

These are just a few factors to pour some cool water into your veins. If AM wins the race towards second generation chips for mass market, you're looking at a 20-25% yield if AM keeps on average 25% of the network consistently over five years.
 
.b