There is a new and probably important paper by Marc J. Melitz and Stephen J. Redding on this topic. For this piece I find a segment in the middle to be more illuminating than the abstract:
Trade has a fractal-like property in this model, in which there are gains from trade at each intermediate stage of production. If one falsely assumes a single stage of production, when production is in fact sequential, these gains from trade at each intermediate stage show up as an endogenous increase in measured domestic productivity. As the number of production stages converges towards infinity, the welfare gains from trade become arbitrarily large. This captures the idea that trade involves myriad changes in the organization of production throughout the economy and the welfare costs from forgoing this pervasive specialization can be large.
As the domestic trade share for an individual production stage becomes arbitrarily small, the welfare gains from trade also becomes arbitrarily large. This captures the idea that some countries may have strong comparative advantages in some stages of production and the welfare losses from forgoing this specialization can be large.
Here is an ungated version of the paper. I would note this idea holds out the hope of integrating the technology diffusion literature with the more traditional international trade theory approaches.