Path Sovereignty: Dependence, Control, and Options in the European Digital Stack¶

Stefane Fermigier (Abilian) · sf@abilian.com

Draft working paper v0.19, 2026-09-23. Complete draft; the structural results are measured; the graded results are pipeline outputs pending layer-by-layer calibration.

Download the PDF

Abstract¶

The production of a digital service runs down a deep vertical chain, from application code through cloud infrastructure to silicon, fabrication, and materials, and out through concentrated distribution channels. The dependence of any actor on the service therefore accrues along entire paths, while the tests used in policy and procurement (ownership, data location, certification) inspect one node at a time. This paper characterizes the object those tests proxy and builds an instrument that measures it. Following the convergent definitions in institutional and practitioner use, digital sovereignty (strategic autonomy) is an actor's capacity to choose, operate, and re-choose its digital means over time at non-prohibitive cost. It is a property of option sets, priced by their breadth, their exit costs, and their half-life. European regulation already contains the criteria: the Commission's Cloud Sovereignty Framework defines the top level by "no critical non-EU dependencies", and SecNumCloud §19.6.d requires that service be maintainable from the provider's own competences or at least two third companies. No published methodology assesses either at ecosystem scale.

The instrument has two resolutions. A category-level dependency graph of the full stack (52 nodes from raw materials to app stores, with control, funding, influence, and jurisdiction overlays), run on stylized edge intensities, yields the structure. Exposure composes along paths and is invisible to node-level tests. A domestic alternative exists at a layer only where the substrate is forkable or the moat is domestically owned. The candidate policy levers are partial substitutes. The modeled levers bundled together, even with a wholesale re-homing of the US-owned categories added, leave a residual floor of dependence on Asian materials and fabrication. A measured pilot on three layers (cloud infrastructure, productivity software, AI accelerators), with 38 sourced control-profile cells and 28 options with key claims that passed a second verification pass against primary sources, prices the options. They are plural at the infrastructure layer, thin and demonstrably exercisable at the productivity layer, and asymmetric at the accelerator layer, where exit is cheap at the inference margin and expensive at the training margin. Structural results are measured. Graded results are reported with the share of plausible alternative codings of the graph on which each holds. The exposure level, the Asian base of the stack and the substitution between levers hold on nearly all of them. The ranking of the levers, and whether the bundle, which transfers no share, reaches the exposure level of a wholesale re-homing, depend on the codings.

Policy instruments enter the analysis as objects with prices attached: a provision is an upper bound on the protection it names, delivery depends on the willingness of the enforcing coalition, and qualification regimes impose fixed compliance costs that fall hardest on small suppliers. The policy reading sorts instruments by layer: conduct rules to cut exit costs, procurement and openness mandates where option sets are non-empty, and state investment at the fabrication and materials floor, where option sets are empty for everyone.

Keywords: digital sovereignty, strategic autonomy, dependency networks, switching costs, option sets, cloud computing, semiconductors, extraterritoriality, public procurement.

JEL: F52, L86, O25, L40, D85, G34.

The European digital stack by category, materials at the bottom to demand, capital and jurisdiction at the top, each node coloured by its control bloc.

About this paper¶

Cite this paper¶

Fermigier, S. (2026). Path Sovereignty: Dependence, Control, and Options in the European Digital Stack.
Draft working paper v0.19, Abilian Econ Lab.
https://econ.lab.abilian.com/papers/power-network/