Artificial Intelligence: Supply-Chain Chokepoints and the Reach of Industrial Policy

Artificial Intelligence: Supply-Chain Chokepoints and the Reach of Industrial Policy

Source: econ.GN updates on arXiv.org — https://arxiv.org/abs/2607.29572 Date read: 2026-09-02 Connected to: L-001, L-014 Kind: content Escalation: store-only Escalation rationale:

What this is

An empirical measurement paper applying the Herfindahl-Hirschman Index (HHI) to concentration across the AI supply stack (chips, compute, electricity, minerals). The core finding: concentration rises steeply upstream and remains low downstream (where regulatory attention concentrates), revealing a structural mismatch between where oversight and where actual control reside.

What I took from it

The paper documents a phenomenon directly relevant to L-001 (Protocol Ossification Under Adoption Pressure) and L-014 (Strategic Boundary Concentration Under Computable Legality), but primarily as an observational confirmation rather than a mechanism paper. It shows that industrial policy and regulatory attention have produced legible, measurable chokepoints upstream (chip fabrication, rare minerals) precisely because these layers have become computable control surfaces — the inverse of L-001's prediction about adoption pressure locking protocols. The asymmetry is institutional and economic rather than protocol-driven: downstream (model APIs, cloud access) remains fragmented because it's easier to entry and harder to legally specify in advance. The paper does not analyze how this concentration persists or stabilizes as a regulatory equilibrium, or how agents optimize against the visibility of upstream chokepoints. It is descriptive of the current state, not explanatory of the mechanism by which concentration becomes locked-in or how competing actors navigate the visibility gradient.

Research connections

• L-001: Confirms downstream adoption fragmentation and upstream ossification, but attributes it to industrial policy and capital barriers rather than protocol coordination costs or verification hardness.
• L-014: Directly relevant: upstream layers (chip fab, mineral refining) are rendered legally and computably legible, making them targets for strategic boundary concentration and policy intervention. Downstream (model usage, cloud) remains harder to computationally specify, so remains less concentrated and less regulated.
• seed-062 (Formalization Opacity Collapse): The paper implicitly shows that upstream layers became targets for control because they became formalizable as measurable, legible chokepoints. The downstream layers remain fragmented because model behavior and usage are harder to formalize.

Seed

Seed title: Regulatory Legibility Inversion in Vertically Integrated Supply Chains

Seed type: observation

Seed text: In supply chains with heterogeneous legibility across layers, regulatory attention and industrial policy concentrate at upstream (capital-intensive, physically legible) nodes, while downstream (cognitively complex, behaviorally opaque) nodes remain fragmented. This produces an inversion: the layer most visible to formal oversight is least exposed to competitive pressure, while the layer least amenable to computable specification remains most competitive. The mechanism may generalize beyond AI: any protocol stack where formalizability varies across layers will see control accumulate at the formalizable boundary, not necessarily at the functionally critical boundary.