We present a worst-case approach for topology optimization under load-uncertainty based on a general problem formulation involving maxima of quadratic functions of an uncertain load vector as both objective and constraints. The problem is reformulated as a non-linear semi-definite program which can be solved efficiently. An important special case of the general problem formulation is worst-case compliance minimization under worst-case stress constraints which is illustrated by numerical examples. (C) 2017 Elsevier B.V. All rights reserved.