01 · Compute is a chokepoint
Unlike ideas or code, compute is physical, countable and made by a few suppliers. That makes it unusually governable, and it means that whoever controls it holds leverage over everyone who needs it.
source-assertionConfidence: highContestability: medium
Compute is an unusually effective governance intervention point because it is detectable, excludable and quantifiable and comes from a concentrated supply chain; compute governance can also threaten privacy, economic outcomes, and centralize power.
Our reading: A preprint position, not a settled finding. Its authors include people at organizations that would operate or be subject to compute governance.
Inspect evidence · 1 passage
Passage summaries are paraphrases, not quotations. Open the source to read the full passage.
Computing Power and the Governance of Artificial Intelligence (opens in a new tab)Abstract
Relative to data and algorithms, AI-relevant compute is presented as a particularly effective point of intervention because it is detectable, excludable and quantifiable and is produced via an extremely concentrated supply chain. Policymakers could use it for regulatory visibility, resource allocation and enforcement; the authors warn that naive or poorly scoped compute governance carries significant risks to privacy, economic impacts and centralization of power, and suggest guardrails.
Scope note: A preprint by authors affiliated with AI developers, universities, and governance organizations; not peer reviewed. Several authors have institutional stakes in compute governance.
Passage checked 2026-09-06 · sastry-compute-abstract
source-assertionConfidence: highContestability: medium
Because only a few firms run their own cloud and compute at scale, much of the wider AI ecosystem depends structurally on them.
Inspect evidence · 1 passage
Passage summaries are paraphrases, not quotations. Open the source to read the full passage.
2023 Landscape: Confronting Tech Power (opens in a new tab)Executive summary, “Computing Power Advantage” (about p. 7 of the PDF)
Building large AI systems depends on cloud and compute that only a handful of firms run themselves, so other AI companies rely on those incumbents.
Scope note: Civil-society voice. An advocacy report with an explicit position on concentration in the technology sector.
Passage checked 2026-10-09 · ai-now-compute-advantage
02 · The divide is geographic
Regional and national strategies from Africa and Latin America describe the same problem from the other side: a lack of high-performance compute, reliance on models and data built elsewhere, and little say in the rules. Distribution is not only about who benefits from AI, but about who can make it.
source-assertionConfidence: highContestability: low
The African Union treats Africa’s lack of compute, and reliance on AI and data developed elsewhere, as a structural barrier and a risk to sovereignty.
Inspect evidence · 1 passage
Passage summaries are paraphrases, not quotations. Open the source to read the full passage.
Continental Artificial Intelligence Strategy (opens in a new tab)Section 1.4.4, p. 28; Executive summary, p. 4
The strategy names the lack of high-performance compute as a main barrier for Africa, notes that all the world’s supercomputers sit in about 30 countries, and that training data usually come from developed countries.
Scope note: Global South regional body. Adopted by the AU Executive Council at its session of 18–19 July 2024 and posted by the AU on 9 August 2024; a strategy, not binding law.
Passage checked 2026-10-09 · au-compute-barrier
source-assertionConfidence: highContestability: low
Chile’s national AI policy says Latin America has been under-represented in setting international AI principles and lacks sovereign high-performance compute.
Inspect evidence · 1 passage
Passage summaries are paraphrases, not quotations. Open the source to read the full passage.
Política Nacional de Inteligencia Artificial (Actualización 2024) (opens in a new tab)Section 3.2, “Articulación Internacional”, p. 11; Section I, p. 3
Emerging countries, especially in Latin America and the Caribbean, have had little part in setting AI principles, a discussion concentrated in the Global North; the region also lacks sovereign high-performance compute.
Scope note: Adopted by decree; published in the Diario Oficial. Quotes are in Spanish; the passage summary is a translation.
Passage checked 2026-10-09 · chile-global-north-participation
03 · Public answers: access, not only rules
Some governments respond to concentration by building or subsidising shared capacity rather than writing new law. International bodies propose a floor under the divide. These are infrastructure interventions: they change who can participate before any safeguard is needed.
source-assertionConfidence: highContestability: low
India answers compute scarcity with public money: its national AI mission funds a public-private compute pool of 10,000 or more GPUs.
Inspect evidence · 1 passage
Passage summaries are paraphrases, not quotations. Open the source to read the full passage.
Cabinet approves over Rs 10,300 crore for IndiaAI Mission (opens in a new tab)IndiaAI Compute Capacity, first of seven pillars
The Cabinet approved a national AI mission with an outlay of Rs 10,371.92 crore, whose first pillar builds public-private AI compute of 10,000 or more GPUs.
Scope note: Official release of the Union Cabinet decision. Current GPU counts are reported elsewhere and change over time.
Passage checked 2026-10-09 · indiaai-compute-pillar
source-assertionConfidence: highContestability: low
India’s AI governance guidelines treat global concentration of compute, data and models as a problem to answer with subsidised access rather than a new AI-specific law.
Inspect evidence · 1 passage
Passage summaries are paraphrases, not quotations. Open the source to read the full passage.
India AI Governance Guidelines: Enabling Safe and Trusted AI Innovation (opens in a new tab)Foreword, p. 2; section 2.1, pp. 14–15
The guidelines note that compute, data and models are concentrated among a few global players, and respond by expanding subsidised access to compute and data rather than with a new AI law.
Scope note: Non-binding guidelines; they state that a separate AI law is not needed at present.
Passage checked 2026-10-09 · india-resource-concentration
source-assertionConfidence: highContestability: medium
Access to compute, data and talent determines who can participate in AI; the UN advisory body proposes a floor under the AI divide rather than universal access to frontier compute.
Our reading: The 'floor' is a proposal, not an existing arrangement.
Inspect evidence · 1 passage
Passage summaries are paraphrases, not quotations. Open the source to read the full passage.
Governing AI for Humanity: Final Report (opens in a new tab)Executive summary, section C 'Common benefits', paras. xxxviii–xl (p. 14)
Effective participation depends on access to talent, compute and data. Compute is described as one of the biggest barriers to entry; the report says it is unrealistic to promise access to compute that even the wealthiest countries and companies struggle to acquire, and instead seeks to put a floor under the AI divide.
Scope note: An advisory body report representing a majority consensus of its members in their personal capacity. It is not a UN member-state decision. Page locators refer to the PDF's printed page numbers.
Passage checked 2026-09-06 · un-hlab-compute-floor
04 · Infrastructure has a physical footprint
Data centres draw electricity and water, and their growth is concentrated in particular places. An infrastructure choice is also a choice about local grids, prices and land, and about which communities carry those costs.
source-assertionConfidence: highContestability: low
Data centres consumed roughly 1.5% of global electricity in 2024, and their consumption is projected to more than double by 2030, concentrated in the United States, China and Europe.
Our reading: The 2024 figure is an estimate and the 2030 figure a base-case projection. Data centres serve more than AI; the report attributes the growth surge to AI.
Inspect evidence · 1 passage
Passage summaries are paraphrases, not quotations. Open the source to read the full passage.
Energy and AI (opens in a new tab)Executive summary — data centre electricity consumption and regional shares
Data centres accounted for around 1.5% of world electricity consumption in 2024, about 415 TWh, growing around 12% a year since 2017. Consumption is projected to more than double to around 945 TWh by 2030. The United States accounted for 45% of 2024 consumption, China 25% and Europe 15%; nearly half of US capacity sits in five regional clusters.
Scope note: Projections are scenario-based estimates, not observations; the 2024 consumption figure is an estimate published in 2025.
Passage checked 2026-09-06 · iea-datacentre-electricity
05 · Worked example · fictional
A research council chooses between cloud credits and a shared cluster
Fictional exercise. A national research council in a lower-income country has a five-year budget for AI research. It can buy credits from a large foreign cloud provider, available immediately, or co-fund a smaller shared cluster with neighbouring countries, ready in two years. No real country, provider or programme is assumed.
1. Who controls the infrastructure under each option?
With credits, the provider sets prices, terms and availability, and can change them. With a shared cluster, control sits with the participating institutions, along with the burden of running it.
2. What becomes hard to reverse?
Five years on one provider's tools, formats and pricing builds skills and data pipelines that fit only that provider. Ask which option keeps the next choice open.
3. Whose data and languages get served?
Compute is not the only input. A cluster without local-language data reproduces the same gap more slowly. Check whether the plan also funds data that reflects the people the research is meant to serve.
4. What does it cost locally?
A domestic cluster needs reliable electricity and cooling. Ask who pays for the grid capacity and whether the energy competes with other needs.
5. Who decides, and on whose behalf?
A council decision commits public money for years. Ask whether researchers, universities and the public had a say, and whether the decision can be reviewed when conditions change.
Speed now and dependence later, or capacity later and independence: neither option is neutral. Reading the choice as infrastructure governance shows the power, distribution and lock-in at stake that a pure cost comparison hides.
What this lesson does not settle
This draft lesson has not yet been reviewed by the framework's author or an outside reviewer. Its sources are policy documents and one civil-society report; they state positions and plans, not measured outcomes. Figures on compute, GPUs and electricity change quickly and are dated in each source. Perspectives from data workers, local communities near data centres and smaller research institutions need further research.