We explore swarm robotics scalability, explaining how cooperation–interference tradeoffs and coordination overhead create a concave performance function with an optimal density. Scalability in swarm robotics, defined as maintaining or improving performance with increasing swarm size, is a crucial system capability often ideally pursued through local communication and decentralized control. However, achieving perfect scalability is often hindered by coordination overhead and swarm density, as increasing the number of robots in a fixed operational area generally leads to interference and efficiency losses. Empirical observations frequently show a concave swarm performance function, peaking at an optimal swarm density where cooperation is balanced with resource contention, resulting in various scaling behaviors from linear to retrograde. Computational models from parallel computing, such as Amdahl’s Law, Gustafson’s Law, and Gunther’s Universal Scalability Law (USL), offer frameworks for analyzing these scaling patterns. The SGF (solo, grupo, fermo) model mechanistically links individual robot states to macroscopic swarm performance and can even derive these classical scalability laws as special cases. A critical insight is the prevalence of two-phase performance or bimodal distributions, revealing that operating near optimal density can place a system precariously close to sudden, catastrophic breakdowns due to cusp bifurcations and hysteresis effects, highlighting the importance of designing for robust and online scalability.