<p>The application of Intelligent Reflecting Surface (IRS) in multi-tier computing networks can effectively enhance spectrum utilization and boost network performance. However, traditional spectrum efficiency optimization strategies often involve a trade-off between system performance and network energy consumption, which increases the complexity of the system model and the difficulty of solving the resulting optimization problem. Therefore, it is crucial to consider energy-efficient computation offloading optimization in IRS-assisted multi-tier computing networks. In this paper, an energy efficiency optimization model is established in an IRS-assisted three-tier computing network, and an energy-efficient multi-tier computation offloading problem that aims to jointly optimize the beamforming matrices for both the IRS and second-tier access points is proposed. To address the resulting parameter coupling and non-convexity of the optimization, a sequential planning approach is adopted to decompose the optimization problem into two stages of energy efficiency maximization and delay minimization. Then, an efficient alternating optimization algorithm integrating Dinkelbach’s method with fractional programming is developed to solve the problem. Finally, simulations demonstrate that the proposed method effectively improves system energy efficiency and achieves notable system performance enhancements.</p>