The relationship between the enhancement of the oceanic carbon sink and seawater acidification has not been fully clarified, and there is a lack of long-term, continuous, and high-coverage seawater CO2 partial pressure data. In order to reduce the uncertainty of global ocean carbon sink estimation caused by insufficient data, a stepwise neural network grid data method is proposed to fit the nonlinear relationship of PCO2 parameters. The results show that the average error between the CO2 partial pressure gridded data and the measured data is 10.64 μatm, and the consistency with the fixed-point continuous observation data is high, and the long-term change trend is in good agreement. The global ocean carbon sink intensity has continued to increase since 2009, reaching 2.23 ± 0.26 PgC yr-1 in 2021. This study provides key data support for analyzing the dynamic process of the oceanic carbon sink, which is of great significance for the improvement of climate models and optimization of carbon management strategies.