My usual workout includes a ten mile bike ride, which I usually complete in about 45 minutes with heavy resistance. According to the machine at my gym, I burn about 500 calories, or 2.1 megajoules, in the process. I'm going to assume the machine means kilogram calories, which is what you see on food labels.
Evidently, I'm putting out a bit more than three quarters of a kilowatt. That's a bit more than one horsepower, which is 745.7 watts. This is a bit surprising -- that's not too far shy of what Wikipedia says you'd expect for the first six seconds of a cycle sprint (900 watts). A professional cyclist can hit about two kilowatts during a sprint. So, 775 watts sustained over 45 minutes is not too shabby.
If they had bothered to wire my exercise bike into the grid, LA Fitness would have obtained 581 watt-hours from my efforts. In Pasadena, we pay $0.15 per kilowatt-hour, so I managed to produce a little less than a dime's worth of electricity. If a hundred people did the same workout, which is roughly what I'd expect over the course of a day, we would together generate $8.72. The gym could save that much by switching off the TVs when no one is watching, or turning down the music a little. The electricity you can generate on an exercise bicycle isn't worth the wires that would carry it.
According to the Department of Energy, the average American household uses about 29.2 kilowatt-hours of electricity per day, so you'd have to pedal at the sprinter's pace of 1216 watts, all day, every day, just to keep up with your household use.
My 2.1 megajoules of peadaling is actually a lot of energy. What astonishes me is that even this rather large amount of energy is worth so little.
Update : My friend Chris points out that the bike at the gym is probably reporting some kind of estimate total power, including power dissipated as heat, that it extrapolates from the work you put into the mechanism. He suggests that around 20-25% of the calories you burn are available as work, so that means I am putting somewhere around 155 to 194 watts into the bike. This probably has an error of 50% or worse, since the bike is extrapolating the total power from the mechanical power, and then I'm extrapolating back to the mechanical power from the result. The actual electricity one could generate is more like $0.02 worth.