In high-latitude regions, carbon dioxide (CO2) emissions during the winter represent an important component of the annual ecosystem carbon budget; however, the mechanisms that control the winter CO2 emissions are currently not well understood. It has been suggested that substrate availability from soil labile carbon pools is a main driver of winter CO2 emissions. In ecosystems that are dominated by annual herbaceous plants, much of the biomass produced during the summer is likely to contribute to the soil labile carbon pool through litter fall and root senescence in the autumn. Thus, the summer carbon uptake in the ecosystem may have a significant influence on the subsequent winter CO2 emissions. To test this hypothesis, we conducted a plot-scale shading experiment in a boreal peatland to reduce the gross primary production (GPP) during the growing season. At the growing season peak, vascular plant biomass in the shaded plots was half that in the control plots. During the subsequent winter, the mean CO2 emission rates were 21% lower in the shaded plots than in the control plots. In addition, long-term (2001–2012) eddy covariance data from the same site showed a strong correlation between the GPP (particularly the late summer and autumn GPP) and the subsequent winter net ecosystem CO2 exchange (NEE). In contrast, abiotic factors during the winter could not explain the interannual variation in the cumulative winter NEE. Our study demonstrates the presence of a cross-seasonal link between the growing season biotic processes and winter CO2 emissions, which has important implications for predicting winter CO2 emission dynamics in response to future climate change.