A point \(p\) in \(\mathbb R^d\) is said to dominate another point \(q\) if the difference \(p-q\) has only nonnegative coordinates. The nondominated points in a set of points are called maxima. The interest of studying dominance and maxima is multifold. First, dominance represents one of the most natural partial orders for multidimensional points, and has been widely used in many scientific disciplines [see \textit{W.-M. Chen, H.-K. Hwang} and \textit{T.-H. Tsai}, Discrete Math. Theor. Comput. Sci. 6, 107--122 (2003; Zbl 1036.68124)]. Second, the number of maxima is itself encountered in many applications like analysis of linear programming and of maxima-finding algorithms [see the paper mentioned above and \textit{M. E. Dyer} and \textit{J. Walker}, Asia-Pac. J. Oper. Res. 15, 159--168 (1998; Zbl 0917.90249)]. Finally, not much is known as far as probabilistic properties of the number of the maxima in dimensions higher than two are concerned. Asymptotic estimates for the mean are usually straightforward, but those for the variance are highly nontrivial even in the simplest case of hypercubes [see \textit{Z.-D. Bai, C.-C. Chao, H.-K. Hwang} and \textit{W.-Q. Liang}, Ann. Appl. Probab. 8, 886--895 (1998; Zbl 0941.60021)]. \textit{Y. Baryshnikov} [Probab. Theory Relat. Fields 117, 163--182 (2000; Zbl 0961.60017)] indicated that the number of maxima in hypercubes is asymptotically normally distributed but without complete proof [see also \textit{A. D. Barbour} and \textit{A. Xia}, Adv. Appl. Probab. 33, 727--750 (2001; Zbl 1005.60028)]. The aim of this paper is to i) derive an asymptotic expansion for the variance of the number of maxima in random samples independently and identically distributed in the hypercube \((0,1)^d\) and ii) derive a central limit theorem with convergence rate for the number of maxima. The main trick used in the paper is the log-transformation first suggested by Baryshnikov (loc. cit.). Switching to a Poisson sample size is introduced by Barbour and Xia (loc. cit.).