Think of this problem as working backward from what you saw on the plate to what was actually present in the original tube of culture, one dilution step at a time.
You spread $50\ \mu\text{L} = 0.05$ mL of the diluted sample and got 20 colonies. Each colony traces back to one live bacterial cell that landed on the plate, so the diluted sample itself must have held
\[ \frac{20\ \text{colonies}}{0.05\ \text{mL}} = 400\ \text{viable cells per mL} \]But this 400 cells/mL figure describes the diluted sample, not the original culture. The sample was diluted 100-fold before you plated it, meaning every mL of original culture was spread out to occupy 100 mL of diluted sample. So the original culture is 100 times more concentrated in live cells than what you measured:
\[ 400\ \text{CFU mL}^{-1} \times 100 = 40000\ \text{CFU mL}^{-1} \]Let's summarize:
So the viable cell count of the original bacterial culture is $40000$ CFU mL$^{-1}$.