SpaceX, and the people in this thread, are comparing the vehicle cost to fuel cost, which is kinda cheating. It's not the cost of the fuel that matters, it's the cost of building the vehicle larger to hold that fuel -- and the fuel needed to launch that fuel -- that matters. So let's do the math!
Most data taken from http://spaceflight101.com/spac...
Stage 1: 23 tonnes structure, 400 tonnes fuel
Stage 2: 4 tonnes structure, 93 tonnes fuel
Payload: 13 tonnes
When launching, the first stage burns all 9 engines at full thrust for two and a half minutes. The re-entry burn and landing happen on a single engine, and from eyeballing the videos (including this one that shows the re-entry burn) appear to take about 30 seconds total. Assuming all burns are near full thrust (which is the best way to do it), that means the landing burn takes about (1/9) * (0:30 / 2:30) = 2% of the first stage fuel. Let's double that to 4% to provide a generous safety margin: that works out to about 400 * 0.04 = 16 tons more fuel.
This fuel is carried up to the moment that the second stage separates, so it subtracts from the mass of the second stage. Second stage plus payload weighs 110 tonnes: without the landing fuel, you could have scaled that up to 126 tonnes, a 15% increase.
So, landing the first stage reduces the payload SpaceX can launch, and thus the money they earn, by about 15%. In exchange, they recover about 75% of the cost of the launch hardware. So it's worth doing, even after you subtract off the cost of recovery and refurbishing. Maybe not the game-changer Elon Musk wants it to be, but it's a win.