Diversity–biomass relationships (DBRs) often vary with spatial scale in terrestrial ecosystems, but the mechanisms driving these scale-dependent patterns remain unclear, especially for highly heterogeneous forest ecosystems. This study explores how mutualistic associations between trees and different mycorrhizal fungi, i.e., arbuscular mycorrhizal (AM) vs. ectomycorrhizal (EM) association, modulate scale-dependent DBRs. We hypothesized that in soil-heterogeneous forests with a mixture of AM and EM tree species, (i) AM and EM tree species would respond in contrasting ways (i.e., positively vs. negatively, respectively) to increasing soil fertility, (ii) AM tree dominance would contribute to higher tree diversity and EM tree dominance to greater standing biomass, and that as a result (iii) mycorrhizal associations would exert an overall negative effect on DBRs across spatial scales. To empirically test these hypotheses, we collected detailed tree distribution and soil information (e.g., nitrogen, phosphorus, organic matter, pH) from seven temperate and subtropical AM–EM mixed forest megaplots (16–50 ha). Using a spatial codispersion null model and structural equation modeling, we identified the relationships among AM or EM tree dominance, soil fertility, tree species diversity, and biomass and, thus, DBRs across 0.01- to 1-ha scales. We found the first evidence overall supporting the three aforementioned hypotheses in these AM–EM mixed forests: (i) In most forests, with increasing soil fertility, tree communities changed from EM-dominated to AM-dominated; (ii) increasing AM tree dominance had an overall positive effect on tree diversity and a negative effect on biomass, even after controlling for soil fertility and number of trees. Together, (iii) the changes in mycorrhizal dominance along soil fertility gradients weakened the positive DBR observed at 0.01- to 0.04-ha scales in nearly all forests and drove negative DBRs at 0.25- to 1-ha scales in four out of seven forests. Hence, this study highlights a soil-related mycorrhizal dominance mechanism that could partly explain why, in many natural forests, biodiversity–ecosystem functioning (BEF) relationships shift from positive to negative with increasing spatial scale.