Ultra-low latency, massive connections, and outstanding energy efficiency are the requisites of sixth-generation (6G) wireless networks. For 6G to reach every corner of the globe, terrestrial ultra-dense networks and NTNs, such as LEO satellites, work hand in hand. Energy efficiency and the dynamics of satellite-terrestrial heterogeneity are, however, typically disregarded by traditional routing algorithms. Introducing an adaptive routing architecture that considers energy for ultra-dense 6G networks that utilise satellite-terrestrial linkages. Using metrics like energy usage, congestion, link quality, and switching costs, SDN, cross-layer optimisation, and traffic prediction powered by machine learning choose the optimal paths on the fly. Energy efficiency, end-to-end delay, and packet delivery ratio all saw significant improvements in simulations across a variety of network configurations. For the communication infrastructures of the future, this demonstrates a sustainable and scalable routing method. © 2025 Elsevier B.V., All rights reserved.