Compute Dependence and Infrastructure Autonomy in Small and Medium States
算力依赖风险与中小国家数字基础设施自主性 Not peer reviewed. Chinese policy-analysis manuscript with an English abstract and reproducible illustrative cost calculations. This is not a national field experiment, country-level dataset or causal policy evaluation. The paper examines compute dependence as a conditional risk involving service continuity, exit capability, operational knowledge and governance. It does not establish inevitable colonial domination or make an allegation against any named country or supplier. Ten scholarly and official sources are identified in a claim-scope register. Statistics on data-centre capacity, cloud-region availability, research spending and electricity demand are not treated as interchangeable measures of AI compute. The original deterministic model enumerates 324 explicitly assumed combinations of annual fixed cost, cloud price, incremental local cost, utilization and relative task efficiency. In an illustrative baseline, reducing task efficiency from 1 to 0.75 raises break-even utilization from 53.51% to 77.83%. These inputs are not observed market quotes or calibrated national costs. Grid frequencies are not empirical probabilities. The paper proposes critical-service continuity capacity, regionally governed shared resources, portable commercial cloud access and domestic operational competence, with limitations and a prospective evaluation design. The PDF and editable DOCX represent the same paper. The ZIP contains the original Python reproduction script, assumed scenario grid, results, source register, README and checksums. No raw personal data, email credentials or third-party full texts are included. CC BY 4.0 covers the original article, code and original scenario outputs; it does not relicense the cited third-party sources. Author: YANGZHOU. Affiliation: Genbibao Embodied AI Research Institute, Shenzhen, China. ORCID: https://orcid.org/0009-0005-4736-2097 . Correspondence: [email protected] . The correspondence address does not establish employment at Sichuan University. Funding: The author confirms no external research funding for this study. Competing interests: Institute affiliation and marketing-related business participation are disclosed; the author confirms no additional interests. OpenAI Codex assisted source discovery, evidence organization, model and code development, execution, checks and drafting. AI is not an author. No human-participant research was conducted. Disclosure clarification added on 20 November 2026: The author's supplied biography describes past activities involving QiSense and an intelligent computing-centre project. The author has confirmed that there is no continuing association with those activities. The historical biographical description is not treated as evidence validating the paper's model or policy conclusions. This manuscript is a policy analysis and is not an independent procurement assessment.
Authors
- YANGZHOU
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-20
- DOI
- https://doi.org/10.5281/zenodo.24376019
- Primary Topic
- Security, Politics, and Digital Transformation
- Type
- preprint