We study dark matter that inelastically scatters and de-excites in direct detection experiments, as an interpretation of the CDMS-Si events in light of the recent LUX data. The constraints from LUX and XENON10 require the mass-splitting between the DM excited and de-excited states to be $|\delta| \gtrsim 50$ keV. At the same time, the CDMS-Si data itself do not allow for a consistent DM interpretation for mass splittings larger than $|\delta| \sim $200 keV. We find that a low threshold analysis will be needed to rule out this interpretation of the CDMS-Si events. In a simple model with a kinetically mixed dark photon, we show that the CDMS-Si rate and the thermal relic abundance can both be accommodated.