The Geography of Compute Prices: Location and Contract Basis, Electricity Constraints, and the Limits of Index Hedging
Energy & Compute Infrastructure →Abstract
A standardized GPU rental contract pays against one reference price across delivery locations. How much protection can it offer a supplier committed to one region? We follow the same eight-GPU configuration across 12 Amazon Web Services regions for 245 days. Even when we choose the reference with hindsight and allow each region its own fixed hedge ratio, 66.13% of aggregate regional quote-change variance remains. Jakarta, Tokyo, and Stockholm account for just over half of the residual. Removing all three and refitting leaves 47.37% across the other nine regions. Service terms also change the geography of prices. Two Google Cloud regions with identical on-demand rates switch rank under interruptible service. In a separate holdout, the fitted hyperscaler index hedge reduces variance in the US sample and increases it globally. Wholesale power indicators show little stable short-run association with repricing. Standardized settlement can cover a common price movement while leaving suppliers exposed to where and on what terms they must deliver.
Keywords
JEL: G13, L86, Q41, D47
Cite this paper
Yang, F.; Liu, Y.; Huang, J.; Huang, J. (2026). The Geography of Compute Prices: Location and Contract Basis, Electricity Constraints, and the Limits of Index Hedging. GARCH Press. https://doi.org/10.69057/WP-0134-V1
Version & publication history
Original repository record
Public catalogue record. The manuscript is available to authorized readers.
Added to GARCH Research Papers
Bibliographic record added. This is not a new manuscript release or a new DOI.
Access & rights
Sign in to preview or download the PDF. Access is available to authorized accounts.