A is true because every DFA is already an NFA.
B is false. Determinization may require exponentially many states, up to $2^n$ in the usual subset construction.
C is true using the standard regular-expression-to-NFA construction.
D is false. The standard product construction for union may require up to $n^2$ product states, and there are language pairs where quadratic state complexity is genuinely necessary. So $2n+1$ is not a valid general upper bound.