Chinese chestnut (Castanea mollissima Blume) seed kernels (CCSK) with high quality and quantity of starch has emerged as a potential raw material for food industry, but the molecular regulatory mechanism of starch accumulation in developing CCSK is still unclear. In this study, we firstly analyzed the fruit development, starch accumulation, and microscopic observation of dynamic accumulation of starch granules of developing CCSK from 10 days after flowering (DAF) to 100 DAF, of which six representative CCSK samples (50-100 DAF) were selected for transcriptome sequencing analysis. Approximately 40 million valid reads were obtained, with an average length of 124.95 bp, which were searched against a reference genome, returning 38,146 unigenes (mean size = 1164.19 bp). Using the DESeq method, 1968, 1573, 1187, 1274, and 1494 differentially expressed unigenes were identified at 60:50, 70:60, 80:70, 90:80 and 100:90 DAF, respectively. The relationship between the unigene transcriptional profiles and starch dynamic patterns in developing CCSK was comparatively analyzed, and the specific unigenes encoding for metabolic enzymes (SUSY2, PGM, PGI, GPT, NTT, AGP3, AGP2, GBSS1, SS1, SBE1, SBE2.1, SBE2.2, ISA1, ISA2, ISA3, and PHO) were characterized to be involved potentially in the biosynthesis of G-1-P, ADPG, and starch. Finally, the temporal transcript profiles of genes encoding key enzymes (susy2, pgi2, gpt1, agp2, agp3, gbss1, ss1, sbe1, sbe2.1, sbe2.2, isa1, isa2, isa3, and pho) were validated by quantitative real-time PCR (qRT-PCR). Our findings could help to reveal the molecular regulatory mechanism of starch accumulation in developing CCSK and may also provide potential candidate genes for increasing starch content in Chinese chestnut or other starchy crops.