Aims. In this paper we aim to determine the structure on 100 AU scales of the massive young stellar object W33A, using interferometric observations in the mid-infrared. This emission could be caused by a variety of elements, for example, the inner protostellar envelope, outflow cavity walls, or a dusty or gaseous accretion disk. Methods. We used the Unit Telescopes of the VLT Interferometer in conjunction with the MIDI instrument to obtain spectrally dispersed visibilities in the N-band on 4 baselines with an angular resolution between 25 and 60 milli-arcsec (equivalent to 95 and 228 AU at 3.8 kpc). The visibility spectra and spectral energy distribution were compared to 2D-axi-symmetric dust radiative transfer models with a geometry that includes a rotationally flattened envelope and outflow cavities. We assumed an O 7.5 ZAMS star as the central source, consistent with the observed bolometric luminosity. The observations were compared to models with and without (dusty and gaseous) accretion disks. Results. The visibilities are between 5% and 15%, and the non-spherically symmetric emitting structure has a typical size of 100 AU. A satisfactory model is constructed to reproduce the visibility spectra for each (u, v) point. It fits the N-band flux spectrum, the midinfrared slope, the far-infrared peak, and the (sub)mm regime of the SED. It produces a 350 μm morphology consistent with the observations. Conclusions. The mid-infrared emission of W33A on 100 AU scales is dominated by the irradiated walls of the cavity sculpted by the outflow. The protostellar envelope has an equivalent mass infall rate of 7.5 × 10 −4 M yr −1 , and an outflow opening angle of 2θ = 20 • . The visibilities rule out the presence of any dust disk with total (gas and dust) mass more than 0.01 M . Within the model, this implies a diskṀ acc of less than 1.1 × 10 −7 (α/0.01) M yr −1 , where α is the viscosity of the Shakura-Sunyaev prescription. However, optically thick accretion disks, which are inside the dust sublimation radius, are allowed to accrete at rates equalling the envelope's mass infall rate (up to 10 −3 M yr −1 ) without substantially affecting the visibilities due to the extinction by the extremely massive envelope of W33A.