This paper argues that attempts at state-building in Afghanistan have led to institutions that are not robust. The state institutions and organizations continue to be highly dependent on external resources and technical expertise, and lack of critical mass of people able and willing to maintain them when external support recedes. I contend that Afghanistan may have fallen into a 'capability trap' that can lead to an actual decrease in state capacity in spite of an appearance of progress. This capability trap has been facilitated by four conditions; (i) high expectations on the government without sequencing or prioritization, (ii) more weight on immediate results than on establishing capable institutions, (iii) a limited menu of acceptable options for institutional arrangements, leading to strong pressures for simple 'transplantation', and (iv) a top-down model of implementation. Thinking about state-building thus needs to shift towards helping to structure or guide a process through which the problem-solving capacity of a broader range of actors can be brought to the fore, and more contextually fit models can emerge, that are less reliant on external expertise, resources, and legitimacy.