<p>Column-supported grain silos are vital for food security but are notoriously vulnerable to earthquakes. A robust framework for assessing their seismic risk is lacking, especially one that accounts for the coupled effects of inter-silo interaction in group configurations, stored material dynamics, and the distinct hazard of pulse-like near-fault ground motions. To address this, we developed a high-fidelity finite element modeling approach, with its dynamic characteristics rigorously validated against shaking table experiments and simplified analytical models. A large-scale probabilistic vulnerability assessment was then performed using Incremental Dynamic Analysis (IDA), with the elastic-plastic column drift serving as the primary damage indicator. Our findings establish that ground motion pulsatility and storage level are the primary drivers of seismic risk. Pulse-like motions consistently increase failure probability; at a design-level PGA of 0·4 g, the collapse risk for a fully loaded row silo rises from 46% (non-pulse) to 52% (pulse). The fully loaded state is unequivocally the most hazardous condition due to amplified inertial forces and dynamic material pressures. Interestingly, the interconnected nature of row silos presents a state-dependent trade-off, offering marginal performance benefits when empty but exacerbating vulnerability compared to single silos when fully loaded. This research provides a new systematic fragility comparison for single versus group silos, delivering validated models for performance-based design and underscoring the urgent need to revise seismic codes to explicitly account for the severe, quantifiable threats posed by near-fault pulses and operational storage levels.</p>