In this paper we prove that the maximum amplitude of a finite-gap solution to the focusing Nonlinear Schrödinger equation with given spectral bands does not exceed half of the sum of the length of all the bands. This maximum will be attained for certain choices of the initial phases. A similar result is also true for the defocusing Nonlinear Schrödinger equation. 1 arXiv:1601.00875v1 [math-ph] 5 Jan 2016 phenomena. Here we refer to the common understanding of rogue waves as exceptionally tall waves with the amplitude |ψ| 2 ≥ 8|ψ| 2 0 , where |ψ 0 | is the amplitude of the background waves. Several particular types of solutions to the fNLS expressed through elementary functions (Peregrine, Akhmediev and Kuznetsov-Ma breathers, see, for example, [10]) provide, perhaps, the most known examples of the rogue wave solutions. These breathers can be viewed as degenerate limits of the corresponding finite-gap solutions.Therefore, it appears natural to look for rogue waves in the class of finite-gap solutions to fNLS. This problem is the subject of an ongoing research [4] for finite-gap solutions of any genus (which is the number of the spectral bands minus one). The main goal of the present paper is a new simple formula for the maximal amplitude of a finite-gap solution with given spectral bands. Namely, we proved that the maximal amplitude cannot exceed half of the sum of the length of all the spectral bands, and this maximum will be attained for certain choices of the initial phases. In fact, due to ergodic property of quasi-periodic solutions, this maximum will be approached by a finite-gap solution with a given spectral bands and generic initial phases in a sufficiently large space-time region. In the case of genus two, this result was recently obtained by O. Wright in [21]. It turns out that the obtained formula is also valid for finite-gap solutions of the defocusing NLS (dNLS) and that a somewhat similar statement is valid for KdV. It will be convenient to describe the finite-gap solutions through the corresponding Riemann-Hilbert Problems (RHPs). It is well known that the inverse scattering transform (IST) method of solving nonlinear integrable systems can be reduced to certain matrix RHP (see [17], [16], [22]), where the jump matrices are defined in terms of the scattering data. The RHPs with permutation type piece-wise constant jump matrices correspond to finite-gap solutions ([6], [5]). In the context of the semiclassical (small dispersion limit) analysis, such RHPs (known as model RHPs or outer parametrices) were first studied in [7] for the KdV and in [18], [15] for the fNLS. They represent the leading order term of the original RHP. Model RHPs are usually obtained through the nonlinear steepest descent method of Deift and Zhou. The finite-gap solution of a model problem provides the local (in x, t) leading order behavior (in the semiclassical limit) of the corresponding slowly modulated solution.Description of results. The data that characterize a finite-gap solution is: (a) a hyperelliptic Riemann surface R...