We develop a field theoretical approach to the cold interstellar medium (ISM) and large structure of the universe. We show that a non-relativistic self-gravitating gas in thermal equilibrium with variable number of atoms or fragments is exactly equivalent to a field theory of a single scalar field φ( x) with exponential self-interaction. We analyze this field theory perturbatively and non-perturbatively through the renormalization group approach. We show scaling behaviour (critical) for a continuous range of the temperature and of the other physical parameters. We derive in this framework the scaling relation M (R) ∼ R d H for the mass on a region of size R, and ∆v ∼ R q for the velocity dispersion where q = 1 2 (d H − 1). For the density-density correlations we find a power-law behaviour for large distances ∼ | r 1 − r 2 | 2d H −6 . The fractal dimension d H turns to be related with the critical exponent ν of the correlation length by d H = 1/ν. The renormalization group approach for a single component scalar field in three dimensions states that the long-distance critical behaviour may be governed by the (nonperturbative) Ising fixed point. The Ising values of the scaling exponents are ν = 0.631..., d H = 1.585... and q = 0.293.... Mean field theory yields for the scaling exponents ν = 1/2, d H = 2 and q = 1/2. Both the Ising and the mean field values are compatible with the present ISM observational data: 1.4 ≤ d H ≤ 2, 0.3 ≤ q ≤ 0.6 . As typical in critical phenomena, the scaling behaviour and critical exponents of the ISM can be obtained without dwelling into the dynamical (time-dependent) behaviour. We develop a field theoretical approach to the galaxy distribution. We consider a gas of selfgravitating masses on the FRW background, in quasi-thermal equilibrium. We show that it exhibits scaling behaviour by renormalization group methods. The galaxy correlations are first computed assuming homogeneity for very large scales and then without assuming homogeneity. In the first case we find ξ(r) ≡< ρ( r 0 )ρ( r 0 + r) > / < ρ > 2 −1 ∼ r −γ , with γ = 2. In the second case we find D(r) =< ρ( r 0 )ρ( r 0 + r) > ∼ r −Γ with Γ = 1. While the universe