Color Glass Condensate (CGC) provides a classical description of dense gluon matter at high energies. Using the McLerran-Venugopalan (MV) model we calculate the initial energy density ε(τ ) in the early stage of the relativistic nucleus-nucleus collision. Our analytical formula reproduces the quantitative results from lattice discretized simulations and leads to an estimate ε(τ = 0.1 fm) = 40 ∼ 50 GeV · fm −3 in the Au-Au collision at RHIC energy. We then formulate instability with respect to soft fluctuations that violate boost invariance inherent in hard CGC backgrounds. We find unstable modes arising, which is attributed to ensemble average over the initial CGC fields.In the relativistic heavy-ion collision Color Glass Condensate (CGC) describes the initial state of energetic gluon matter with the transverse momentum p t up to the saturation scale Q s which universally characterizes the hadron or nucleus wavefunction in the smallx regime [1,2,3,4,5,6,7,8,9,10]. Given the scale Q s at a certain value of Bjorken's x, the gluon distribution probed by processes with Q 2 ≪ Q 2 s is so dense that coherent fields should be more relevant than the individual particle picture during τ Q −1s . Physically Q 2 s corresponds to the transverse density of partons and is estimated by the Golec-Biernat and Wüsthoff fit,, where A is the atomic number. We can expect Q s around 1 ∼ 2 GeV for the Relativistic Heavy Ion Collider (RHIC) and 2 ∼ 3 GeV for the Large Hadron Collider (LHC) in case of A = 197 (Au-Au collision) assuming relevant p t is ∼ 1 GeV. This transient but still coherent gluon matter, which is often referred to as "Glasma" [7], should melt toward a quark-gluon plasma.The physical property of Glasma has been mainly analyzed by numerical simulations in the lattice discretized formulation [4,5,6,7,8,9]. In this paper we aim to approach Glasma in an analytical way along a similar line to the near-field expansion proposed by Fries-KapustaLi [10]. The analytical method is desirable for a deeper insight into the Glasma, which presumably exists up to τ Q −1 s ∼ 0.1 fm in the Au-Au (central) collision at RHIC energy, √ s = 200 GeV/nucleon, or even longer depending on the interpretation of the Glasma. In particular, the problem of early thermalization still has interesting unanswered questions [11]. We will specifically address the following; is there any unstable mode growing around the initial CGC fields right after the collision? If any, it could speed up thermalization (or isotropization) even in the classical regime (τ Q −1 s ), besides non-Abelian plasma instabilities [12,13,14,15] which take place at later times. The pioneering numerical simulation [9] suggests the existence of "Glasma instability", though the literal time scale of instability seems to be greater than Q −1 s by three order of magnitude, probably because of the choice of tiny instability seeds. The delay in the instability onset has also been pointed out in the Hard Expanding Loop (HEL) approach to non-Abelian plasma instabilities [15].We shall start w...