The Measure Space is a living research platform, policy engine, and global organizing hub commited to the cause of Planetary Majority-Public Compute.
FOUNDING FRAMEWORK · DRAFT v0.1 · 2026
Prepared by: Alexander Wilson for The Measure Space
For comment and coalition review. Not final.
Executive Summary
The most concerning concentration of power today is not a government or military, but a relatively small group of private individuals who control the computational substrate of future civilization: the semiconductor fabrication plants, the data centers, the AI models, the orbital compute and telecom constellations, at a scale that dwarfs the capacities of any democratic public.
Our founding framework presents four interlocking components:
•The Enclosure Research Programme: a continuously updated, open-access atlas of AI power concentration across various analytical layers, in particular: physical compute infrastructure, ownership and capital flows, governance power and legal legitimacy, elemental extractions and flows of natural ressources, space infrastructure.
•The Mandatory Public Compute Contribution: a structural policy proposal aiming to ensure that the future possibilties of planet belong to the inhabitants of the planet: legal instruments that require that for every unit of compute manufactured, a majority contribution is made to a publicly governed compute commons. Private ambition becomes the engine of public capacity. 50% + 1 of compute operating on the surface or in orbit of our planet must be public.
•The Harm Proximity Principle: a new foundation for international institutional legitimacy: governance authority over shared infrastructure is proportional to exposure to the harms that infrastructure can produce, and inversely proportional to ownership of it.
•The Global Governance Body: an institutional design for a multi-stakeholder commons governance body whose composition is maintained by an open-source, formally verified, fully auditable and transparent protocol, resistant to capture and ossification by design.
1.1 A Threshold Has Been Crossed
Preliminary estimates, based on current hyperscaler capital expenditure trajectories, suggest that individual private compute portfolios in the range of $50–100 billion in asset value begin to approach or exceed the total AI infrastructure capacity of most democratic governments. Several actors have already crossed this threshold. The rate of divergence is accelerating.
1.2 Why This Moment Is Different From All Previous Inequality
The compute-capital feedback loop has four properties that distinguish it from all prior forms of wealth concentration:
Self-amplification: Capital has alway been self-amplifying, but in contact with AI this aspect is supercharged. More compute produces better AI. Better AI generates more wealth and power. More wealth buys more compute. The loop has no natural ceiling.
Displacement of human cooperation: Previous wealth concentration still required masses of workers, consumers, and voters to function. Automated systems increasingly do not. This leads to a collapse of the "social contract" by which a population voluntarily concedes some of its sovereignty to government in exchange for security and stability.
Speed asymmetry: A well-resourced AI system can model regulatory responses, identify legal vulnerabilities, simulate public opinion, and optimize lobbying strategies faster than any legislature can respond. Democracy runs on human time. This runs on machine time.
Infrastructural invisibility: With natural resource extraction, people can see the mines and the pipelines. When a private actor controls the AI layer of a hospital system, a welfare payment system, or a military logistics network, the control is not visible to those subject to it.
1.3 The Wealth-Compute Flywheel
The enclosure of common land in 16th–18th century England systematically transferred shared resources into private ownership, dispossessing communities that had depended on them for survival. The intelligence enclosure of the 21st century transfers the infrastructure of intelligence, once notionally public in its scientific foundations and collaborative academic origins, into private hands at unprecendented and exponentially accelerating rates.
Compute is the only economic resource that serves as the substrate on which governance, warfare, healthcare, education, financial systems, and democratic deliberation itself will now run. To allow its majority ownership to remain in private hands is to allow the privatization of the preconditions of public life. It is tantamount to abandoning the project of human civilization.
1.4 The Current Landscape
The following data represents available evidence on compute concentration. Significant gaps in public data exist and are documented in Section 6.
Semiconductor Fabrication
•Taiwan Semiconductor Manufacturing Company (TSMC) fabricates approximately 80% of advanced semiconductor nodes globally. A single facility in a geopolitically contested territory constitutes a critical single point of failure for global AI development.
•High Bandwidth Memory (HBM), essential for training and running large language and reasoning models is supplied by three firms: SK Hynix (~62% market share), Micron (~21%), and Samsung (~17%).
•Nvidia commands approximately 92% of data center AI GPU sales, protected by the CUDA developer ecosystem, creating deep vendor lock-in across research and commercial AI.
•China controls approximately 60% of critical mineral processing capacity required for advanced chips and energy storage.
Cloud and Compute Infrastructure
•Amazon (AWS), Microsoft (Azure), and Google (GCP) together hold an estimated 65–70% of the global cloud infrastructure market, which includes the primary market for GPU compute used in AI research and deployment.
•This infrastructure is geographically concentrated. Amazon's highest-performance GPU instances are housed in approximately seven global locations; in most countries, AI researchers must route data and workloads through foreign-controlled infrastructure.
•The six leading firms are projected to consume between 239–295 TWh of electricity annually by 2030, approximately 1% of total global power demand, concentrated in specific regional grids in North America, Western Europe, and Asia-Pacific.
Foundation Models
•The ability to train frontier AI models is effectively limited to a handful of actors with access to sufficient compute capital. The training compute required for leading models has increased approximately 4x per year since 2012.
•OpenAI, Google DeepMind, Anthropic, Meta AI, and xAI constitute the de facto frontier of foundation model development, all based in the United States.
Ownership and Capital
•The five largest technology companies by market capitalization collectively hold more financial assets than the majority of sovereign states. SOURCE NEEDED: Updated comparative analysis of corporate treasury capacity versus national AI infrastructure budgets.
•Pension funds and sovereign wealth funds representing the savings of hundreds of millions of ordinary people are among the largest investors in the infrastructure of compute concentration.
•Cross-shareholding and board interlocks among the leading AI infrastructure firms create coordination patterns that are not visible through standard market analysis.