The greenhouse gas (GHG) emissions from fuel-based lighting are substantial given the paltry levels of lighting service provided to users, leading to a great opportunity for GHG mitigation by encouraging the switch from fuel-based to rechargeable LED lighting. However, as with most new energy technology, switching to efficient lighting requires an up-front investment of energy (and GHGs) embedded in the manufacture of replacement components. We studied a population of off-grid lighting users in 2008-2009 in Kenya who were given the opportunity to adopt LED lighting. Based on their use patterns with the LED lights and the levels of kerosene offset we observed, we found that the embodied energy of the LED lamp was paid for in only one month for grid charged products and two months for solar charged products. Furthermore, the energyreturn- on investment-ratio (energy produced or offset over the product's service life divided by energy embedded) for off-grid LED lighting ranges from 12 to 24, which is on par with on-grid solar and large-scale wind energy. We also found that the energy embodied in the manufacture of a typical hurricane lantern is about one-half to one-sixth of that embodied in the particular LED lights that we evaluated, indicating that the energy payback time would be moderately faster if LEDs ultimately displace the production of kerosene lanterns. As LED products improve, we anticipate longer service lives and more successful displacement of kerosene lighting, both of which will speed the already rapid recovery of embodied energy in these products. Our study provides a detailed appendix with embodied energy values for a variety of components used to construct off-grid LED lighting, which can be used to analyze other products.