Terrestrial ecosystems assimilate atmospheric carbon dioxide (CO2) through photosynthesis, which provides carbon substrates that fuel a wide range of organisms and accounts for about a third of the annual global carbon sink. The devastating impacts of extreme weather and climate change have urged the need to reduce greenhouse gas emissions and evaluate long-term changes in ecosystem structure and function. Modeling climate change impacts on terrestrial ecosystems remain uncertain, as the land surface modules used in Earth System Models (ESMs) often do not reproduce the observed ecosystem responses to experimental manipulations (e.g., atmospheric CO2 enrichment). Here, we employ a biogeochemical model (ecosys) that reasonably represents observed microbe-soil-plant-atmosphere interactions to estimate changes in water and carbon cycling across 69 global eddy covariance sites (eight ecosystem types) under the high-emissions climate warming scenario (SSP5-8.5). Our simulations show primarily positive trends over the 69 sites for aboveground biomass (93% of sites), net primary productivity (90% of sites), and water-use efficiency (94% of sites) by 2100, suggesting that CO2 fertilization could partly offset anthropogenic increases in CO2 emissions. Importantly, our analysis indicates large carbon cycle sensitivities to belowground water and nutrient availability modulated by root exudates and soil water potential, whose parameterizations are currently uncertain in ESMs. Our model framework recognizes observational constraints when quantifying ecosystem responses to climate change, which has the potential to refine assessments of nature-based climate solutions.