Governing the Compute Frontier: Rival Regimes for Artificial Intelligence Infrastructure and the Case for a Third Path

Abstract

The governance of artificial intelligence is usually described as a single global project, yet the three powers that lead in the field have built three different regimes for governing the infrastructure on which artificial intelligence depends. This article characterizes those regimes and asks what each treats as the governable object, what all three systematically neglect, and what a viable alternative for states outside the triad would require. The method is a qualitative, comparative, and doctrinal analysis of primary policy instruments and authoritative secondary analysis across the United States, the European Union, and China, read against the compute-governance literature. The analysis finds that the United States governs through market-led construction coupled to security-driven export control, the European Union through a threshold-triggered law of general-purpose models and a standards apparatus, and China through the state-coordinated construction of national computing capacity. Each regime is optimized for its own strategic position, and each is blind to the same two things: the energy and climate layer on which all compute runs, and the agency of the states that will host much of the physical infrastructure. The article argues that a governable compute frontier requires a third path defined by four criteria: governing the physical and energy layer rather than only the model, defaulting to interoperability and transparency, treating compute access as a development question, and distributing governance authority beyond the three powers. The implications concern global policy, the competition among the powers, and the design of workable governance for the majority of states.

Keywords: ai governance, compute, digital infrastructure, geopolitics, energy policy, global south

DOI: 10.7176/ISDE/16-1-04

Publication date: September 30^th 2026