In desert soils, water vapor transport is a key mechanism driving soil moisture dynamics. This transport occurs along vapor density gradients that arise from variations in soil matric potential and temperature in the soil. In hyper-arid soils, extremely high negative matric potentials across the near surface lead to uniform matric potential gradients (∇Ψₘ). Therefore, in these soils, variations in temperature (∇T) primarily drive vapor movement by affecting vapor concentration. In cases where soil moisture is high enough so that the relative humidity in the air-filled pores approaches 100%, vapor flows from warmer to cooler soil sections. This is because an increase in temperature under these conditions causes an exponential rise in vapor concentration, which triggers vapor migration toward cooler soil regions. We hypothesize that in very dry soils, the opposite is the case. When the relative humidity in the air-filled pores is much lower than 100%, an increase in temperature does not translate to an increase in water vapor because there is no liquid water to evaporate. In contrast, the increase in temperature results in a decrease in air density, forming a lower water vapor concentration compared to a cooler soil, resulting in water vapor flow from cooler to warmer soil sections. To test this hypothesis, we conducted an in-situ experiment in the Negev Desert, Israel, where the total soil-atmosphere water flux was measured by lysimeters, and the soil water content at depths of 0.5, 2, 5 and 10 cm were measured using temperature and relative humidity sensors. Water vapor transport was also simulated using a HYDRUS 1D numerical model. We found that vapor transport in these hyper-arid soils is dominated by thermally driven vapor flux (total soil water flux ≈ thermal vapor flux), while liquid fluxes (thermal and isothermal) and isothermal vapor fluxes are negligible. While the experimental data support our hypothesis, the HYDRUS configuration does not allow for an influx of water vapor from the atmosphere, nor does it allow for water vapor to move from cooler to warmer soil layers, both of which may limit the model’s prediction accuracy. These results highlight the need to reconsider the description of water vapor flow in extremely dry soils in HYDRUS and potentially other land-surface and hydrological models.