The Measure Space

3. The Compute Enclosure Research Programme

The Compute Enclosure Research Programme is the evidence engine of The Measure Space. It is a continuously updated, open-access, rigorously sourced database and visualization platform mapping AI power concentration across several analytical layers. It is designed to become the reference document that every AI governance negotiation must reckon with.

Unlike existing work, which is fragmented across disciplines, static in format, US/EU-centric in perspective, and absent of any explicit intervention-point analysis, the Compute Enclosure Research Programme integrates the full vertical stack of AI power into a single coherent framework, updated continuously, with Global South perspectives structurally represented.

Layer 1: Physical Compute Infrastructure

The physical substrate of AI power: where it is, who owns it, and what its vulnerabilities are.

•

Semiconductor fabrication plant locations, ownership, capacity by node size, and geopolitical exposure

•

Hyperscale and AI-specific data center locations, power consumption, water usage, and ownership chains

•

Undersea cable networks: routes, ownership, landing points, and documented interference incidents

•

Energy infrastructure dependencies per compute cluster: grid connections, energy sources, vulnerability to disruption

•

Satellite internet and edge compute constellations: orbital positions, coverage, ownership, and governance voids

Primary data sources: DC Maps, TeleGeography, ITU, company annual reports, satellite imagery analysis, national energy regulators, academic geography literature.

Critical data gaps: No comprehensive public database of AI-specific (vs general cloud) data center capacity by owner exists. Energy consumption figures are largely self-reported. Undersea cable ownership structures are partially obscured through subsidiary arrangements. Real-time compute utilization data is entirely proprietary.

Layer 2: Ownership and Capital

The financial architecture of AI power: who controls what, how capital flows concentrate it further, and which non-obvious actors are implicated.

•

Corporate ownership trees for the top 50 AI infrastructure actors: subsidiaries, holding companies, cross-shareholdings

•

Sovereign wealth fund and pension fund exposure to AI infrastructure equity, the hidden connection between ordinary people's savings and the infrastructure of their own disempowerment

•

Key directorial board interlocks and personal networks among AI infrastructure leadership

•

Revolving door mapping: movements between AI companies, regulatory bodies, and government positions

•

Government contracts with hyperscalers: value, duration, scope, and documented conflicts of interest

Primary data sources: SEC and equivalent national regulatory filings, OpenCorporates, Orbis/Bureau van Dijk database, parliamentary and congressional lobbying registers, government procurement databases.

Critical data gaps: Cross-shareholding structures are poorly captured in standard financial databases. Pension fund exposure to AI infrastructure is distributed across complex fund-of-fund structures that resist easy aggregation. Personal network and informal relationship data requires original research. Non-US revolving door data is largely uncollected.

Layer 3: Governance Power

What decisions does controlling each node actually confer? This is the layer most absent from existing work.

•

Documented cases of service disconnection, throttling, or deprioritization (AWS and Parler, Cloudflare and content moderation decisions, payment processor deplatforming)

•

Which nodes operate with no democratic oversight, no public accountability mechanism, and no appeal process?

•

AI system deployment in public sector services: welfare, policing, border control, healthcare: mapping where algorithmic governance has replaced democratic accountability

•

Military and intelligence infrastructure dependencies on private AI providers

Primary data sources: Terms of service and acceptable use policy analysis, documented deplatforming incidents, government AI deployment disclosures, civil liberties organization case databases, Freedom of Information requests.

Critical data gaps: Military AI contracts are largely classified. Government AI deployment in the Global South is almost entirely undocumented. The full scope of intelligence agency compute dependency on private infrastructure is unknown. No systematic database of "algorithmic governance incidents", cases where AI systems have made consequential public decisions without meaningful oversight, currently exists.

Layer 4: Dependency Mapping

Who is locked in, how deeply, and at what cost to their sovereignty?

•

National AI infrastructure dependency ratios: proportion of national compute capacity that is privately foreign-owned versus domestically public

•

Institutional lock-in: universities, hospitals, government agencies, and militaries whose core operations depend on single-vendor AI infrastructure

•

Global South digital infrastructure dependency: mapping the nations with effectively zero sovereign compute capacity and the consequences for their policy autonomy

•

Switching cost analysis: what would it cost, in time and resources, for a given institution or nation to exit its current primary AI infrastructure dependency?

Primary data sources: World Bank digital infrastructure data, ITU connectivity statistics, national AI strategy documents (where they exist), academic research on digital sovereignty, civil society organization reports from Global South.

Critical data gaps: No comprehensive global mapping of institutional-level AI infrastructure dependency exists. Switching cost data is entirely absent from public literature. Global South compute sovereignty data is severely underrepresented in all existing research, including this project's current evidence base.

Layer 5: Intervention Opportunities

Where does democratic leverage exist? This layer inverts the standard analytical frame: instead of describing power concentration as a problem to be studied, it maps it as set of pragmatic affordances.

•

Existing regulatory handles with demonstrated or potential efficacy: antitrust instruments, export controls, utility designation, procurement standards, data localization requirements

•

Where public alternatives exist or are being built: national compute initiatives, academic federated compute networks, public cloud infrastructure projects

•

Historical precedents where comparable concentrations were successfully constrained: AT&T monopoly breakup, spectrum public interest obligations, Basel capital requirements, nuclear non-proliferation framework

•

Current active campaigns: which organizations are challenging which nodes, using which instruments, with what current status

•

Coalition mapping: who needs to be in the same room to make a specific intervention viable?

Primary data sources: Regulatory filing databases, legislative tracking services, civil society organization networks, comparative public utility law literature, international law and treaty databases.

Critical data gaps: No systematic mapping of active democratic interventions on AI infrastructure concentration exists. Coalition mapping is an almost entirely original research task. The legal feasibility of various intervention instruments across different jurisdictions is poorly understood and requires original legal analysis.