The distribution of assimilated carbon among the plant parts has a profound effect on plant growth, and at a larger scale, on terrestrial biogeochemistry. Although important progress has been made in modelling photosynthesis, less effort has been spent on understanding the carbon allocation, especially at large spatial scales. Whereas several individual-level models of plant growth include an allocation scheme, most global terrestrial models still assume constant allocation of net primary production (NPP) among plant parts, without any environmental coupling. Here, we use the CASA biosphere model as a platform for exploring a new global allocation scheme that estimates allocation of photosynthesis products among leaves, stems, and roots depending on resource availability. The philosophy underlying the model is that allocation patterns result from evolved responses that adjust carbon investments to facilitate capture of the most limiting resources, i.e. light, water, and mineral nitrogen. In addition, we allow allocation of NPP to vary in response to changes in atmospheric CO2. The relative magnitudes of changes in NPP and resource-use efficiency control the response of root:shoot allocation. For ambient CO2, the model produces realistic changes in above-ground allocation along productivity gradients. In comparison to the CASA standard estimate using fixed allocation ratios, the new allocation scheme tends to favour root allocation, leading to a 10% lower global biomass. Elevated CO2, which alters the balance between growth and available resources, generally leads to reduced water stress and consequently, decreased root:shoot ratio. The major exception is forest ecosystems, where increased nitrogen stress induces a larger root allocation.