Three-dimensional (3D) neural cell cultures inherently lend themselves to high-throughput network electrophysiology studies addressing brain function in health and disease in a more realistic architectural complexity than two-dimensional neural networks. Epilepsy is the emblem of brain network disorders, as it reflects aberrant circuit reorganization and hyper-synchronization, resulting in sudden and uncontrolled electrical discharges (seizures). Modeling the features of epilepsy has so far relied on pharmacological, ionic or genetic manipulation of cells, ex-vivo brain tissue or intact animals, failing to recapitulate most of the epilepsies, which are triggered by unknown causes. Here, we report the spontaneous emergence of epileptiform patterns in spheroids of rodent primary hippocampal cells cultured in physiological condition, i.e., in the absence of a known initiating insult, detected by microelectrode array electrophysiology. Three distinct electrical phenotypes, i.e. interictal (between seizures), ictal (seizure) or mixed, arise from DIV10 to DIV35. In particular, the tonic-clonic ictal discharges become the most prominent at DIV28-35. These patterns exhibit electrographic and spectral features that strikingly resemble those observed in the hippocampus of in vitro and in vivo rodent epilepsy models, as well as of drug-resistant epileptic humans. Remarkably, not all spheroids exhibit full-blown ictal activity, bringing parallelism with the yet unanswered question of why a brain becomes epileptic and a seizure is generated. This evidence warrants caution against hippocampal cell-based therapies for regenerative purposes, as they may initiate epileptogenesis; at the same time, hippocampal spheroids lend themselves as reductionist model supporting high-throughput pre-clinical research on epileptic syndromes involving the hippocampus.