This paper presents a comprehensive survey of Data-Efficient Graph Learning (DEGL), a research area aimed at improving graph machine learning performance with limited labeled data. The paper introduces DEGL as a new research frontier and provides a systematic review of recent advances in three key areas: self-supervised graph learning, semi-supervised graph learning, and few-shot graph learning. It highlights the challenges of training models with large labeled data and discusses the importance of data-efficient methods in practical scenarios where labeled data is scarce. The survey also identifies promising directions for future research, emphasizing the need for methods that can operate effectively with limited labeled data. Self-supervised graph learning focuses on learning representations without relying on external labels. It includes methods like generation-based, contrastive-based, and auxiliary property-based approaches. These methods aim to learn meaningful representations by leveraging the structure and relationships within graph data. Semi-supervised graph learning combines labeled and unlabeled data to train models. It includes label propagation, consistency regularization, and pseudo-labeling methods. These approaches aim to improve model performance by utilizing the relationships within the data. Few-shot graph learning focuses on learning models that can generalize with a small number of labeled examples. It includes metric learning and parameter optimization approaches. These methods aim to transfer prior knowledge across different tasks with high generalization capacity. The paper discusses the challenges and opportunities in data-efficient graph learning, including robustness, generalizability, and the integration of large models. It also highlights the need for further research in extending graph learning to non-Euclidean spaces and providing theoretical justifications for data-efficient learning. The survey concludes with a discussion of future research directions and the potential impact of data-efficient graph learning on real-world applications.This paper presents a comprehensive survey of Data-Efficient Graph Learning (DEGL), a research area aimed at improving graph machine learning performance with limited labeled data. The paper introduces DEGL as a new research frontier and provides a systematic review of recent advances in three key areas: self-supervised graph learning, semi-supervised graph learning, and few-shot graph learning. It highlights the challenges of training models with large labeled data and discusses the importance of data-efficient methods in practical scenarios where labeled data is scarce. The survey also identifies promising directions for future research, emphasizing the need for methods that can operate effectively with limited labeled data. Self-supervised graph learning focuses on learning representations without relying on external labels. It includes methods like generation-based, contrastive-based, and auxiliary property-based approaches. These methods aim to learn meaningful representations by leveraging the structure and relationships within graph data. Semi-supervised graph learning combines labeled and unlabeled data to train models. It includes label propagation, consistency regularization, and pseudo-labeling methods. These approaches aim to improve model performance by utilizing the relationships within the data. Few-shot graph learning focuses on learning models that can generalize with a small number of labeled examples. It includes metric learning and parameter optimization approaches. These methods aim to transfer prior knowledge across different tasks with high generalization capacity. The paper discusses the challenges and opportunities in data-efficient graph learning, including robustness, generalizability, and the integration of large models. It also highlights the need for further research in extending graph learning to non-Euclidean spaces and providing theoretical justifications for data-efficient learning. The survey concludes with a discussion of future research directions and the potential impact of data-efficient graph learning on real-world applications.