Objective: In recent years, China's high-speed train industry has developed rapidly. As a key core component, the process of localizing axle box bearings is limited by insufficient fatigue performance. At present, high-speed train bearings rely entirely on imports, and fatigue of domestically manufactured bearing rings has become an urgent problem that must be addressed. The fatigue performance of axle box bearing rings is mainly determined by the metallurgical quality of the bearing steel and the heat treatment process. G20CrNi2MoA steel is a high-quality carburized bearing steel for manufacturing the inner and outer rings of this type of bearing. Although domestic smelting levels have gradually caught up with international levels, there is still a significant gap in carburizing heat treatment technology, and the hardness distribution of the carburized layer is the key factor determining the rolling contact fatigue life of carburized steel bearings. To address this dilemma, the outer ring of the domestic G20CrNi2MoA bearing is the core research object of this study, aiming to explore the influence of the hardness distribution of the carburized layer on the bearing's rolling contact fatigue performance and to optimize the process. Methods: A systematic research system of experimental verification, simulation, and intelligent optimization was constructed. First, the key factors affecting performance were determined through a rolling contact fatigue test, after which the formation mechanism of the carburized layer was analyzed via simulated heat treatment. Subsequently, a model combining a genetic algorithm (GA) and a backpropagation (BP) neural network was introduced to optimize the parameters. Results: Using this system, innovative results were achieved. The results confirm that a carburized layer with a moderate depth is key to extending the fatigue life of the outer ring, and the model provides a more accurate solution for the hardness distribution curve of the optimal carburized layer (surface hardness: 693 HV; depth of carburized layer: 1.71 mm). Based on this reverse optimization, a complete carburized heat treatment scheme is obtained. The key parameters include a long infiltration time of 16.5 h, a diffusion temperature of 930 ℃, a diffusion time of 6.54 h, a carbon diffusion potential of 1.05%, and an isothermal time of 3.6 h. Targeted tests verify that the rolling contact fatigue life of the outer ring of the bearing treated using this process is approximately 4.7% higher than that of the existing domestic bearing outer ring. Conclusions: First, there is a significant nonlinear correlation between the rolling contact fatigue performance of the domestic G20CrNi2MoA bearing outer ring and the hardness distribution of the carburized layer, the depth of the carburized layer, and other parameters. Accurately regulating these parameters is key to improving performance. Second, the model combining a GA and a BP neural network provides an efficient and accurate technical approach for optimizing the carburizing heat treatment of bearings, thereby greatly reducing the cost and cycle time of traditional trial-and-error methods. Third, the carburizing heat treatment process proposed in this study effectively improves the fatigue performance of domestic bearing rings. It provides important support, with both theoretical value and practical guiding significance, for addressing the technical gap in the heat treatment of high-speed bearings in China, breaking the monopoly of foreign technology, and promoting the localization of high-speed train bearings. It also lays a foundation for subsequent research and the development of higher-speed grade bearings.