Sovereign Digital Infrastructure Series (SDIS) — SDIS 05: Compute Sovereignty

SDIS-05 — Compute Sovereignty: AI Infrastructure, Semiconductor Dependency, and Strategic Autonomy in the GPU Era

Compute sovereignty — guaranteed, uninterruptible access to the processors that run a nation's digital and AI infrastructure — has become the most capital-intensive and geopolitically exposed layer of the sovereignty stack introduced in SDIS-01. This paper examines the structural dependency created by extreme concentration in advanced semiconductor fabrication and AI accelerator supply, and traces the history of compute's transformation from a commodity input into an explicitly strategic asset.

The paper surveys the national compute strategies emerging in response — domestic GPU cluster procurement, sovereign cloud compute allocation, and diversified accelerator sourcing — and provides a three-dimensional framework for estimating a nation or institution's compute exposure. It examines fourteen detailed case studies spanning successful diversification programs, single-vendor dependency incidents, firmware licensing outages, and comparative national strategies across the Gulf states, India, and the EU, addresses eight common objections, and closes by connecting compute sovereignty to the operational and confidential-computing controls in SDIS-02 and the resilience testing methodology in SDIS-07, alongside an implementation roadmap, glossary, and self-assessment checklist.

Keywords: sovereign AI, compute capacity, GPU supply chain, national AI stacks, export controls

This is Paper 5 of 12 in the Sovereign Digital Infrastructure Series (D-AI), extending the four-plane framework in SDIS-01 into the compute plane and connecting directly to the sovereign cloud patterns in SDIS-02 and the resilience methodology in SDIS-07. A companion series guide (SDIS-00) provides a full cross-citation map, publication registry, and licensing terms for all papers.

Series: Sovereign Digital Infrastructure Series (SDIS) · Publication: SDIS-05 · Author: Bijan Burnard · Affiliation: D-AI Research · Version: 1.0 · Publication Date: 23 August 2026 · ISBN: 978-969-9597-33-6 · DOI: 10.5281/zenodo.22083821 · License: CC BY 4.0