Realistic large-scale networks display an heterogeneous distribution of connectivity weights, that might also randomly vary in time. We show that depending on the level of heterogeneity in the connectivity coefficients, different qualitative macroscopic and microscopic regimes emerge. We evidence in particular generic transitions from stationary to perfectly periodic phase-locked regimes as the disorder parameter is increased, both in a simple model treated analytically and in a biologically relevant network made of excitable cells. (i) the precise number of receptors and the extremely slow plasticity mechanisms induce a static disorder termed quenched synaptic heterogeneities.(ii) Thermal noise, channel noise and the intrinsically noisy mechanisms of release and binding of neurotransmitter [4] result in stochastic variations of the synaptic weights termed stochastic synaptic noise. Experimental studies of cortical areas [5] showed that the degree of heterogeneity in the connections significantly impacts the input-output function, rhythmicity and synchrony. Yet, qualitative effects induced by connection heterogeneities are still poorly understood theoretically. One notable exception is the work of Sompolinsky and collaborators [6]. In the thermodynamic limit of a onepopulation firing-rate neuronal network with synaptic weights modeled as centered independent Gaussian random variables, they evidenced a phase transition between a stationary and a chaotic regime as the disorder is increased. Besides stationary and chaotic regimes, synchronized oscillations is a highly relevant macroscopic state. It is related to fundamental cortical functions such as memory, attention, sleep and consciousness, and its impairments relate to serious pathologies such as epilepsy or Parkinson's disease [7,8].In this Letter, we show that both quenched and stochastic heterogeneities induce the emergence of macroscopic, perfectly periodic oscillations in the meanfield limit, corresponding to phase-locked oscillations of all neurons in the network. A generic transition from a stationary to a periodic regime as disorder increases will first be demonstrated in the case of firing-rate neurons. Similar reasoning will lead us to uncover the same transition in a more realistic network featuring excitable cells.Models of cortical areas involve at least two neural populations, one excitatory corresponding e.g. to pyramidal neurons and one inhibitory population modeling interneurons. In the firing-rate model with N neurons and P populations, the dynamics of the membrane potential (V i , i ∈ {1 · · · N }) of all neurons in the network is given by the system equations:where p i is the population index of neuron i, τ pi is a common characteristic time of all neurons of population p i and I pi (t) their deterministic input. The interaction with the other neurons in the network is given by the sum of a synaptic efficacy J ij multiplied by the firing rate of the presynaptic cell, a positive sigmoidal transform (S(·)) of the voltage. Network heterogeneiti...