This paper analyzes the influence of the long-run decline in US energy intensity on projections of energy use and carbon emissions to the year 2050. We build on our own recent work which decomposes changes in the aggregate US energy–GDP ratio into shifts in sectoral composition (structural change) and adjustments in the energy demand of individual industries (intensity change), and identifies the impact on the latter of price-induced substitution of variable inputs, shifts in the composition of capital and embodied and disembodied technical progress. We employ a recursive-dynamic computable general equilibrium (CGE) model of the US economy to analyze the implications of these findings for future energy use and carbon emissions. Comparison of the simulation results against projections of historical trends in GDP, energy use and emissions reveals that the range of values for the rate of autonomous energy efficiency improvement (AEEI) conventionally used in CGE models is consistent with the effects of structural changes at the sub-sector level, rather than disembodied technological change. Even so, our results suggest that US emissions may well grow faster in the future than in the recent past.