Directed acyclic graphs (DAGs) are crucial in hardware synthesis and compiler optimization. Synthetic DAGs can be used for benchmarking computing systems while preserving intellectual property. However, DAG generation is challenging due to the inherent directional and logical dependencies. This paper introduces LayerDAG, an autoregressive diffusion model. LayerDAG decouples the strong dependencies into units that can be processed sequentially. By interpreting the partial order of nodes as a sequence of bipartite graphs, LayerDAG leverages autoregressive generation to model directional dependencies and employs diffusion models to capture logical dependencies within each bipartite graph. Experiments demonstrate that LayerDAG outperforms existing DAG generative models, particularly for generating large-scale real-world DAGs with up to 400 nodes.