Neuromorphic Computing for AGI: A Systematic Review Beyond GPU Limitations

Abstract

Intelligence (AI) has achieved remarkable progress in recent decades, mainly driven by parallel computing architectures such as Graphics Processing Units (GPUs). However, as AI models grow in complexity and energy demand, GPUs show fundamental limitations in scalability and biological plausibility, restricting further progress toward Artificial General Intelligence (AGI). In contrast, neuromorphic computing reproduces the dynamics of biological neural systems through event-driven and massively parallel architectures, offering ultra-low-power processing and adaptive learning capabilities. This paper presents a Systematic Literature Review (SLR) following PRISMA 2020 and PICOC frameworks to identify how SpiNNaker, a brain-inspired architecture from the University of Manchester, contributes to AGI research when compared with GPUs. The review considers studies published between 2024 and 2025 retrieved from Scopus, resulting in 36 primary sources after applying inclusion and exclusion criteria. Findings show that SpiNNaker and related neuromorphic chips demonstrate energy-efficiency improvements of up to two orders of magnitude, enhanced scalability, and real-time adaptive behavior. These results confirm that neuromorphic hardware represents a feasible path toward sustainable and biologically plausible AGI systems.