Dendritic Spikes as a Model for Biophysically Grounded Computation in Neuromorphic Systems

This study synthesizes recent advances in our understanding of dendritic computation—particularly the role of NMDA receptor-mediated dendritic spikes—and evaluates how these biologically grounded mechanisms are currently being approximated in neuromorphic hardware systems. Building on a growing body of experimental work combining two-photon calcium imaging with dendritic patch-clamp electrophysiology, we highlight how clustered synaptic activation can evoke local regenerative events (NMDA spikes) with amplitudes exceeding 40 mV and durations in the hundreds of milliseconds. These findings suggest that dendritic branches function as nonlinear subunits capable of complex spatiotemporal integration, challenging traditional point-neuron abstractions such as leaky integrate-and-fire models.The methodological component of this work consists of a critical comparative analysis between biological mechanisms and engineering implementations, with a focus on how current neuromorphic systems attempt to replicate (or fundamentally diverge from) these dendritic computations. We reviewed hardware platforms including IBM’s NorthPole chip (notable for its 25× energy reduction through memory-compute co-location), Intel’s Hala Point architecture (featuring 1.15 billion neurons across 140,000+ cores), and various emerging memristive technologies that exhibit device-intrinsic plasticity.This work does not involve new experimental data or human/animal subjects; therefore, ethical approval or data collection compliance was not applicable. However, all referenced experimental studies discussed were drawn from peer-reviewed sources that report adherence to institutional ethical standards for animal research. All hardware data and performance benchmarks were obtained from published technical documentation, conference proceedings, and peer-reviewed articles where available.By integrating insights from experimental neuroscience, computational modeling, and hardware design, this study aims to provide a conceptual and practical roadmap for implementing biologically plausible, energy-efficient, and scalable neuromorphic systems. We conclude by identifying critical biological features—such as structural plasticity, graded dendritic integration, and multi-frequency oscillatory coupling—that remain under explored in hardware and may hold the key to future advances in artificial general intelligence.