It's all about labor costs. To set up a second shift on the same line there are two options:
1. Use the same people from first shift, but give them overtime. But this still isn't enough man power, so you need to hire temp labor. Overtime + temp labor is very, very pricey. No way to reach 25% margin with this approach.
2. Hire another shift worth of people. But this is basically doubling the workforce and requires a massive amount of work to hire and train people. I would imagine you keep 1/2 your people on shift #1 and move 1/2 your people to shift #2. Then, you'd hire 1/2 more new people for shift #1 and 1/2 more new people for shift #2. It would take some time for the new people to get up to speed. And there's no easy way to do 1/2 shift. It's either 1 shift (with 1 shift of workers) or 2 shifts (with 2 shifts of workers). To do 1.5 shifts is extremely difficult and I think it would require the 1 shift of workers plus overtime and temp labor.
All in all, it's easiest to just focus on 1 shift on 1 line and max out efficiency. Once you reach 25% gross margin with this, then you launch your second shift on the same line by basically doubling your workforce. But then your volume output jumps from 20k units a year to 40k units a year. Thus, this is why Elon has been forecasting 20k units (now 21k) in 2013 and 35k+ in 2014.
Lastly, my hunch is that the 20k units for one shift (one line) isn't exactly fixed. I think it's a bit flexible and they probably do a bit more on one shift and keep 25% margin (maybe up to 24k annual run-rate on one shift?).