Ziegler, Micah S. and Jessica E. Trancik

Lithium-ion technologies are increasingly employed to electrify transportation and provide stationary energy storage for electrical grids, and as such their development has garnered much attention. However, their deployment is still relatively limited, and their broader adoption will depend on their potential for cost reduction and performance improvement. Understanding this potential can inform critical climate change mitigation strategies, including public policies and technology development efforts. However, many existing models of past cost decline, which often serve as starting points for forecasting models, rely on limited data series and measures of technological progress. Here we systematically collect, harmonize, and combine various data series of price, market size, research and development, and performance of lithium-ion technologies. We then develop representative series for these measures, while separating cylindrical cells from all types of cells. For both, we find that the energy capacity–scaled real price of lithium-ion cells has declined by about 97% since their commercial introduction in 1991. Using performance curve models, we estimate that between 1992 and 2016, real price per energy capacity declined 13% per year for both all types of cells and cylindrical cells, and upon doubling cumulative market size, decreased 20% for all types of cells and 24% for cylindrical cells. We also develop a method to incorporate additional performance characteristics into these models, including energy density and specific energy performance metrics. When energy density is incorporated into the definition of service provided by a lithium-ion battery, estimated technological improvement rates increase considerably. The annual decline in real price per service increases from 13 to 17% for both all types of cells and cylindrical cells while learning rates increased from 20 to 27% for all cell shapes and 24 to 31% for cylindrical cells. These increases suggest that previously reported improvement rates might underestimate the rate of lithium-ion technologies’ change. Moreover, our improvement rate estimates suggest the degree to which lithium-ion technologies’ price decline might have been limited by performance requirements other than cost per energy capacity. These rates also suggest that battery technologies developed for stationary applications, where restrictions on volume and mass are relaxed, might achieve faster cost declines, though engineering-based mechanistic cost modeling is required to further characterize this potential. The methods employed to collect these data and estimate improvement rates are designed to serve as a blueprint for how to work with sparse data when making consequential measurements and forecasts of technological change.