A point source on a plane constantly emits particles which rapidly diffuse and then stick to a growing cluster. The growth probability of a cluster is presented as a sum over all possible scenarios leading to the same final shape. The classical point for the action, defined as a minus logarithm of the growth probability, describes the most probable scenario and reproduces the Laplacian growth equation, which embraces numerous fundamental free boundary dynamics in non-equilibrium physics. For non-classical scenarios we introduce virtual point sources, in which presence the action becomes the Kullback-Leibler entropy. Strikingly, this entropy is shown to be the sum of electrostatic energies of layers grown per elementary time unit. Hence the growth probability of the presented non-equilibrium process obeys the Gibbs-Boltzmann statistics, which, as a rule, is not applied out from equilibrium. Each layer's probability is expressed as a product of simple factors in an auxiliary complex plane after a properly chosen conformal map. The action at this plane is a sum of Robin functions, which solve the Liouville equation. At the end we establish connections of our theory with the tau-function of the integrable Toda hierarchy and with the Liouville theory for non-critical quantum strings. The goal of this work is to unify two fundamental highly non-equilibrium processes, Laplacian growth (LG) [1][2][3][4][5], which is deterministic interface dynamics, and diffusion-limited aggregation (DLA) [6] -a discrete universal stochastic fractal growth. These remarkable processes have a lot in common and were suspected to be deeply related [7][8][9][10][11].Laplacian growth raised enormous interest in physics because of (i) its impressively wide applicability ranging from solidification and oil recovery to biological growth [1], (ii) remarkable universal asymptotic shapes, it exhibits [1,[12][13][14][15][16], and (iii) discoveries of deep intriguing connections of LG to quantum gravity [2] and the quantum Hall effect [17]. In mathematics the Laplacian growth appears so exciting because it possesses beautiful and powerful properties, unusual for most of nonlinear PDEs, such as infinitely many conservation laws [18] and closed form exact solutions [16,[19][20][21][22]. A new splash of intense activity in LG (see [3] for a review) was provoked by the work [2], where strong connections of LG with major integrable hierarchies and the theory of random matrices were established.Mathematical formulation of LG is (deceptively) simple: a droplet of air, D + (t), where t is time, is surrounded by a viscous fluid, D − (t) = C/D + (t), called D(t) for simplicity. Both liquids are sandwiched between two parallel close plates. Fluid velocity in D(t) obeys the Darcy law, v = −∇p (in scaled units), where p(z,z) is pressure and z = x + iy is a complex coordinate on the plane. Because of incompressibility, ∇·v = 0, then ∇ 2 p = 0 in D, except points with sources, which provide growth. Also, p = 0 at the interface, Γ(t) = ∂D(t), between two fluids, if...