Because prediction of the load-carrying capacity of piles continues to challenge geotechnical engineers, new solutions are needed. The problem is aggravated by the lack of understanding of the phenomena of soil-pile interaction, and the limited quantity and inexact quality of subsurface soil information that can be provided for analysis. The use of fuzzy set theory improves the engineer's ability to handle the uncertainty in the soil parameters and the prediction methods and thus improves the reliability of the predicted capacity. A methodology for predicting pile capacity based on fuzzy set theory is developed and implemented in a computer program. The validity of the program results is evaluated using a US Federal Highway Administration (FHWA) pile load test database. The comparison between predicted and measured ultimate capacities of piles in sand, clay and mixed soils shows that the predictive capability of the developed program is very good.