That's the expected behavior: the brackets are symmetric with respect to the formula axis, which is the imaginary line where the main fraction line sits on.
I guess that this is some step during dimensional analysis; suppose you follow the advice by Bernard, but put it in context:
\documentclass{article}
\usepackage{amsmath}
\begin{document}
\[
\begin{bmatrix}
\dfrac{1}{\dfrac{\mathrm{L}^3}{\mathrm{t}}}
\end{bmatrix}
=
\left[\frac{\mathrm{t}}{\mathrm{L}^3}\right]
\]
\end{document}
Your readers will have a very hard time in interpreting the meaning of such a formula. Which one is the main fraction line? The top one is just slightly longer than the (secondary) bottom one and L3 lies almost exactly at the same level as the fraction line in the right-hand side.
Take into account that mistaking the main fraction line would lead to very wrong interpretation.
Making the main fraction line longer doesn't really solve the problem:
\[
\begin{bmatrix}
\dfrac{\;\;1\;\;}{\dfrac{\mathrm{L}^3}{\mathrm{t}}}
\end{bmatrix}
=
\left[\frac{\mathrm{t}}{\mathrm{L}^3}\right]
\]
would produce
which is as unreadable.
Solution: use exponents, as is commonly done in dimensional analysis:
\documentclass{article}
\usepackage{amsmath}
\begin{document}
\[
\left[\frac{1}{\mathrm{L}^3\mathrm{t}^{-1}}\right]
=
[\mathrm{L}^{-3}\mathrm{t}]
\]
\end{document}
\left[\frac{1}{\mathrm{L}^3\mathrm{t}^{-1}}\right]
anyway. Or, directly,[\mathrm{L}^{-3}\mathrm{t}]
\left[ \frac{1}{\mathrm{L}^3/\mathrm{t}} \right]
. Much more readable.