Abstract: Aiming at the long-term goal of independent and sustainable development of artificial intelligence(AI) fundamental software, and considering the efficient resource management needs of AI models and applications in intelligent computing scenarios, this article first analyzes the multi-dimensional and whole-process supporting role of intelligent fundamental software for AI tasks from spatial and temporal perspectives. Then, it explores the current status and challenges of building data and algorithm aware intelligent fundamental software. It also presents the key design ideas for such software in terms of resource management, load perception, task scheduling, and large language model support. Finally, it discusses potential future research directions and technologies based on this study, aiming to provide crucial support for the large-scale implementation and continuous innovation of AI.