The effect of nanoparticle aggregation on the thermal conductivity of nanocomposites or nanofluids is typically nonnegligible. A universal model (Maxwell model) including nanoparticle aggregation is modified in order to predict the thermal conductivity of nanocomposites more accurately. The predicted thermal conductivities of silica and titania nanoparticle powders are compared first with that measured by a hot-wire method and then with those in previous experimental works. The results show that there is good agreement between our model and experiments, and that nanoparticle aggregation in a nanocomposite enhances the thermal conductivity greatly and should not be ignored. Because it considers the effect of aggregation, our model is expected to yield precise predictions of the thermal conductivity of composites.