In this study, we analyze structural changes in embodied CO2 trade due to the US withdrawal from the Paris Agreement by fusing CGE and IO analysis. The research method is the following two stage method. In the first stage, we investigate the economic and carbon impacts of the Paris Agreement with and without the US using the GDyn-E model and the GTAP 10 Data Base. In the second stage, we apply I-O analysis to the ex-ante and ex-post simulation interregional I-O tables. I-O analysis shows the interdependence of the CO2 load (embodied CO2 trade). Furthermore, it allows us to decompose the change in CO2 emissions into such contributions as the emission coefficient change, the technology change, and the final-demand change. We found that with the US departure from the Paris Agreement, the world's CO2 emissions in 2030 increase by about 2.4 billion tons compared to the case without the US departure. This is due to an increase in CO2 emitted in the US. However, our I-O analysis reveals that about 8.3% of the increase in CO2 is due to the increase in consumption-based CO2 emissions in countries other than the US. In other words, if a large country like the US leaves the agreement, it will undermine the effectiveness of carbon reduction policies by increasing CO2 imports of other countries. From the structural decomposition analysis, we found that improvement of CO2 output ratio contributes most to carbon reduction. However, the US withdrawal diminishes this contribution in carbon-restricted countries.