AMPL does support conic programming when used with the CPLEX and Gurobi solvers. See for example these slides:
http://www.ampl.com/MEETINGS/TALKS/2012_08_Berlin_Thu.1.H1058.pdf http://www.ampl.com/MEETINGS/TALKS/2013_08_Lisbon_Thu.A.23.pdf
In brief, conic constraints are represented in AMPL as quadratic constraints, with a sum of squares of variables on the left-hand side, and either the square of a nonnegative variable or the product of two nonnegative variables on the right-hand side. Actually it is a little more general, as any term may be multiplied by a positive constant.