Context. The Spitzer spacecraft monitored the Deep Impact event on 2005 July 4 providing unique infrared spectrophotometric data that enabled exploration of comet 9P/Tempel 1’s activity and coma properties prior to and after the collision of the impactor. Aims. The time series of spectra take with the Spitzer Infrared Spectrograph (IRS) show fluorescence emission of the H2O ν2 band at 6.4 μm superimposed on the dust thermal continuum. These data provide constraints on the properties of the dust ejecta cloud (dust size distribution, velocity, and mass), as well as on the water component (origin and mass). Our goal is to determine the dust-to-ice ratio of the material ejected from the impact site. Methods. The temporal evolution of the continuum was analyzed using a dust thermal model which considers amorphous carbon and intimate silicate-carbon mixtures. The water emission was extracted from the spectra and the water columns within the Spitzer extraction aperture were inferred using a fluorescence excitation model. Time-dependent models simulating the development of the ejecta cloud and the sublimation of icy grains were developed to interpret the temporal evolution of both dust and water emissions within the field of view. Results. Both the color temperature of the grains in the ejecta cloud (375 ± 5 K) and their size distribution show that a large quantity of submicron grains were ejected by the impact. The velocity of the smallest grains is 230 m s −1 , with a power index for the size dependence of 0.3−0.5, in agreement with gas loading. The total mass of dust is (0.7−1.2) × 10 5 kg for grain sizes 0.1−1 μ ma nd (0.5−2.1) × 10 6 kg for sizes 0.1−100 μm. A sustained production of water is observed, which can be explained by the sublimation of pure ice grains with sizes less than 1 μm and comprising a mass of ice of (0.8−1.8) × 10 7 kg. The contribution of dirty ice grains to the ice budget is negligible. Assuming that water was essentially produced by icy grains present in the ejecta cloud, our measurements suggest a very high ice-to-dust ratio of about 10 in the excavated material, which greatly exceeds the gas-to-dust production rate ratio of ∼0.5 measured for the background coma. Alternately, if a large amount of material felt back to the surface and sublimated, ice-to-dust ratios of 1−3 are not excluded. A better understanding of the cratering event on 9P/Tempel 1 is required to disentangle between the two hypotheses. Evidence for grain fragmentation in the ejecta cloud is found in the data. The pre-impact water production rate is measured to be (4.7 ± 0.7) × 10 27 mol s −1 .