Feynman integrals are complicated objects and it is generally hard to evaluate them analytically. I will present recent work that showed that these integrals are, in contrast to their analytic evaluation, surprisingly well-suited for numerical integration. While traditional methods are struggling with integrals at the 2-3 loop level, integrals with general Euclidean kinematics and up to 15 loops can be easily evaluated using a novel 'tropical' approach. A key step towards this highly efficient numerical Feynman integration is a certain simplification of the underlying QFT. This simplified QFT, which I call the tropical(ized) QFT, is obtained in a tropical deformation limit of the original QFT.