Aspects of the disclosure relate to using machine learning models to automatically generate spine-leaf network topologies. A computing system may receive one or more prompts to generate a spine-leaf network topology based on a non-spine-leaf network topology of a non-spine-leaf network. Based on inputting the prompts into a natural language processing model, network criteria for generating the spine-leaf network topology may be generated. Non-spine-leaf network topology data comprising network metadata, network dependency parameters, and network constraint parameters, may be retrieved. Based on inputting the network criteria and the non-spine-leaf network data into a generative adversarial network implemented on a quantum computing device, candidate spine-leaf topologies comprising a qualified candidate spine-leaf network topology that meets the network criteria may be generated. A block of a blockchain may be generated and may comprise the qualified candidate spine-leaf network topology that meets the network criteria.