Firm growth shows two robust anomalies. First, volatility falls with size much more slowly than independence would predict: σ(S) ∼ S −β with β ≈ 0.15–0.20 rather than 1/2. Second, the growth-rate distribution is sharply peaked and fat-tailed rather than Gaussian. Standard explanations put the source of these effects inside each firm. We ask instead whether the same facts can emerge between firms, from a network of interacting companies in which no heterogeneity is imposed externally. We model the economy as a Generalized Lotka–Volterra system on a sparse random graph, held at its cooperative critical point µc. This regime is hard to reach numerically: there the total abundance diverges in finite time and breaks explicit solvers.