Context: The carbon dioxide (CO2) flux in a sugarcane ( Saccharum officinarum L.) ecosystem is of great importance when attempting to mitigate global climate change associated with the eco-economic importance of sugarcane for sugar and biofuel production. China is one of the major sugarcane producing country in the world. However, CO2 flux and carbon budget of sugarcane ecosystems in China remains unclear. Objective: The objectives of this study were (1) to observe seasonal and inter-annual variations in the CO2 flux at different time scales; (2) characterize how the CO2 flux responds to environmental factors; and (3) to quantify the carbon budget to assess whether a sugarcane ecosystem is a net CO2 sink after considering the carbon removal in the harvested stalk and N2O emissions. Methods: The eddy covariance techniques combined with soil respiration and N2O emission measurements using a static chamber method were used to observe CO2 fluxes and quantify carbon budget in a representative sugarcane plantation located in South China during three continuous sugarcane growing seasons from 2021 to 2023. Results: Our results showed that the CO2 fluxes exhibited clear and large single-peak inter-diurnal, seasonal and inter-annual variations closely related to dynamics in environmental factors and the sugarcane growth and development stages. The air (Ta) and soil temperature (Ts), net radiation (Rn), photosynthetic photon flux density (PPFD), and leaf area index (LAI) were the important factors controlling CO2 variations, followed by vapor pressure deficit (VPD). Principal component analysis showed that energy and water related factors explained 59.20 % and 25.63 % of the variation in CO2 fluxes, respectively. Photosynthetic parameters of sugarcane ecosystem varied with growth stages attributed to dynamics in biotic factors (such as LAI). The seasonal totals for net ecosystem CO2 exchange (NEE), ecosystem respiration (Reco), gross primary productivity (GPP), soil respiration (Rs), and aboveground autotrophic respiration (Raa) were -991 f 228 g C m-2, 1038 f 69 g C m-2, 2027 f 180 g C m-2, 518 f 51 g C m-2, and 519 f 120 g C m-2, respectively. The net ecosystem carbon balance was -67 f 11 g C m -2 after considering carbon removal in the harvested stalks (925 f 235 g C m-2). Topsoil (0-20 cm) organic carbon content showed increasing trends during 2019-2023 attributed to annual leaf-C (300 g C m-2) and root-C (100 g C m-2) returning. However, the high N2O emissions (3.38 f 1.82 g N m-2) caused by high conventional fertilizer N input (500 kg N ha-1), while much lower N2O emissions (0.97 f 0.25 g N m-2) under reduced fertilizer N input (300 kg N ha-1) were observed. The net global warming potential was much lower under N300 (174 f 141 g CO2 eq. m-2) than N500 (1208 f 792 g CO2 eq. m-2). Conclusions: Energy related factors (such as PPFD, Rn, Ta, and Ts) contributed the most to the changes in CO2 flux during sugarcane growing season. The sugarcane ecosystem in China was an overall net carbon source under conventional fertilizer N input after considering the carbon removal in the harvested cane stalks and high N2O emissions. Implications: The results implied that the sugarcane ecosystem in China had potentials to fix CO2 by optimizing fertilizer N rate combined with enhanced efficiency N fertilizers (such as slow-release fertilizer, nitrification and urease inhibitors) and returning crop residue.