Abstract
The growth of large loads in power systems suggests that markets for data and markets for power will be increasingly interlinked. In this environment, optimizing system welfare requires revelation of the values associated with electricity consumption. This paper introduces the Value of Compute Load (VoCL), a framework for estimating the implied willingness-to-pay for electricity from observable cloud computing prices. Using available public information we construct estimates of the short-run marginal value for over 75 cloud computing instance loads with a focus on accelerator-powered instances relevant for AI applications.
Bottom-up estimates reveal a wide range of estimates for the value of compute load, differentiated by region, by time and by reliability of service. We also derive estimates for the value of flexibility and the value associated with forward contracting of compute load. The implications for market design are material, suggesting a refocus on full strength price formation, combined with appropriate hedging and service prioritization schemes.
Billimoria, Farhad and Conleigh Byers. “Bits-to-Watts: Connecting Markets and Prices for Compute and Power.” United States Association for Energy Economics, July 31, 2026
The full text of this publication is available via United States Association for Energy Economics.