Water scarcity in agriculture, climate variability, and the growing demand for food have generated a pressing requirement for effective irrigation management systems. IoT-driven intelligent irrigation offers a solution to the problem through its combination of field sensing, wireless communication, edge computing, cloud computing, and control. The area has seen little consolidation since current approaches vary greatly in their architecture, hardware, communication, control, and evaluation methods.This review critically examines nine dimensions of IoT-based smart irrigation: system architecture, sensing and edge hardware, communication, edge computing, cloud integration, irrigation intelligence, energy management, performance evaluation, and security. A taxonomy is developed according to the location of sensing, inference, learning, control, and data synchronization across field nodes, edge gateways, and cloud platforms. The literature shows broad convergence toward layered sensor–edge–gateway–cloud architectures, but reliability depends mainly on the placement of decision-making, buffering, and recovery functions. Edge-based intelligent computing improves resilience during connectivity loss, while platforms such as STM32 microcontrollers combined with LoRa or LoRaWAN support energy-efficient local sensing, inference, and autonomous control. However, on-device inference introduces a trade-off between computational energy consumption and savings from reduced radio transmission. Major gaps remain in offline buffering, edge–cloud reconciliation, security, and comprehensive field validation. No reviewed system performs strongly across water savings, energy use, communication reliability, latency, security, scalability, and cost simultaneously.