Whoever owns the AI owns the future.
As AI, data, and compute become the dominant productive force, the question of ownership of this infrastructure is the most important political question of the 21st century.
Where do things stand today, and where do they break?
Today, AI value creation is concentrated in the hands of a few companies. The top seven — Microsoft, Alphabet, Meta, Amazon, Apple, Nvidia, Tesla — account for over 30% of S&P 500 market capitalisation. That is historically unprecedented. Foundation models are controlled by a handful of actors: OpenAI, Anthropic, Google DeepMind, Meta, xAI. A large share of training data comes from public sources whose creators receive no share of the value creation. Compute infrastructure is concentrated among a few hyperscalers. The architecture resembles that of electricity infrastructure, but with far less regulation.
Classical market economics works when value creation is tied to human labour: wages flow to millions. AI decouples value creation from labour and concentrates it with the owners of the models. Without correction, we arrive at a techno-feudal structure: 90% consumers, 10% rent recipients. That is not a market economy — it is a rentier economy.
Paths out of the crisis
The stars show how realistic a path is: 1 star = barely feasible, 5 stars = highly feasible.
Sovereign AI fund (Norway model)
The state secures stakes in foundation models, AI infrastructure, and data pools, along the lines of an oil fund.
+ Democratises value creation through capital rather than regulation.
Only works with early entry.
Trade-off: Political independence versus democratic oversight.
Feasibility: Model is globally established (Sovereign Wealth Funds hold ~$12tn AuM). Broadly accepted politically. Realistic for the next generation of AI and mid-tier champions.
Data Cooperatives
Citizens collectively manage their data (MIDATA Switzerland, established 2015; Salus Coop Barcelona).
+ Federated, citizen-controlled, GDPR/Swiss FADP-compliant.
Scaling challenge against hyperscalers.
Trade-off: Sovereignty at the expense of competitiveness.
Feasibility: A legal framework exists; the GDPR supports it. A growing number of working examples. Politically uncontroversial: nobody lobbies against cooperatives.
Open AI Commons
Foundation models as digital commons (BLOOM, Mistral, EU EuroHPC).
+ Prevents monopolisation.
Compute costs enormous (GPT-4 >$100m training).
Trade-off: Democracy versus speed.
Feasibility: The EU is actively advancing this agenda (EuroHPC, Mistral). Realistic for specific domains (public administration, science, public services); difficult to achieve as a universal competitor to hyperscalers.
Citizen Capital Funds
Mandatory stakes for large companies in citizen funds (Sweden, the Meidner Plan of 1976, never implemented).
+ Ownership vested in citizens without nationalisation.
Politically very difficult; capital flight risk.
Trade-off: Maximum distribution against maximum resistance.
Feasibility: Intellectually elegant, politically utopian outside Scandinavia. Sweden's Meidner Plan was proposed in 1976 and never implemented.
Platform Co-operative
Platforms in the cooperative model (Stocksy, Up & Go NYC, Mondragon).
+ Market-compatible, democratic, scalable.
Capital-access constraint against well-capitalised competitors.
Trade-off: Works in niches; not universally applicable.
Feasibility: A proven model in the cooperative tradition. It works in niches: care, local logistics, creative work. Scaling it against Big Tech is difficult without state support.
Trigger-based phases
Trigger: First G7 or EU country adopts an explicit sovereign AI fund mechanism with a strategic-stake mandate.
Policy is acknowledging that AI ownership is a matter of strategic national interest. The first explicitly sovereign AI funds have emerged in Scandinavia, the Middle East, and Singapore. The EU is discussing a strategic-stake mechanism. Data co-operatives are growing towards 5m–10m members across Europe.
Trigger: sovereign AI fund holds more than 5% in a top-50 AI company for the first time
Sovereign funds hold 10–20% stakes in European AI champions. Platform co-operatives lead the market in 2–3 sectors (mobility, care, local logistics). Open-AI commons become the standard in public administration. Switzerland establishes a participation-fund mechanism.
Trigger: Sovereign wealth funds become the largest institutional buyer group for AI champions globally.
The ownership structure is hybrid: privately controlled, with substantial citizen and state stakes. Returns are broadly distributed. Techno-feudal concentration is averted.
Real-world pilot programmes
Where the ideas behind this pillar are already being tested. "Learn more" opens the explanation without leaving the page.
What we are watching
SWFs are becoming structural buyers of AI champions.
The cooperative model has surpassed critical mass; scaling is viable.
Europe has substantial sovereign compute capacity of its own; Open AI Commons are scaling.
Track in the index →Strategic-stake mechanism is structurally established.
A political backlash is inevitable; debate over Sovereign Stakes is going global.
Track in the index →Our assessment
At Globalance, we regard the question of ownership as the central question of the economy ahead., more important than UBI, more important than a robot tax. Whoever controls capital automatically controls the distribution question in the end. All other modules — taxes, income, democracy — are consequences of the ownership pattern, not causes.
Viable is a Dual Strategy: above one Sovereign AI Fund at the national level, underpinned by a dense network of Data Cooperatives and Platform cooperatives. At the top, value creation is secured through large models and infrastructure; at the bottom, through data, local platforms and sector-specific applications. Both layers together create a breadth of ownership that preserves the market economy without tipping into techno-feudal concentration.
Open AI Commons are important as a European strategy, but realistic only in specific domains: public administration, science, and education. They will not replace hyperscalers, but they can prevent critical public functions from becoming entirely dependent on private infrastructure. EU-EuroHPC and Mistral are the right steps; they need more capital and political backing.
Citizen Capital Funds in the Meidner tradition we consider intellectually elegant, but politically utopian outside Scandinavia. With the AHV, Switzerland has a precursor that is capable of being extended. An 'AHV 2.0' with a substantial capital-stock component would be conceptually feasible and politically realistic. It would in effect constitute a Citizen Capital Fund with generational trust as its custodian.
From a Swiss perspective conditions are favourable: the co-operative tradition is deeply rooted (Migros, Coop, Raiffeisen, MIDATA), sovereign wealth could be legitimised through a link to the AHV, and direct democracy can carry a participation-fund initiative. Switzerland would thus be well positioned to assume a first-mover role in this area.
What this means for investors: First, anyone wishing to share in the value creation of AI needs to own a part of it. Broadly diversified stakes along the entire chain — from infrastructure through models to applications — are the most direct route. Second, public investment funds, such as those channelled through the AHV, could become large, long-term investors and reshape the ownership structure of entire industries. Third, cooperative and platform models are gaining ground where data comes from many sources. They are rarely accessible to investors, yet they put pressure on the margins of established platforms. Fourth, concentration matters: the more narrowly returns are concentrated among a few, the greater the political risk for those very positions.
Why now: We are in a phase where ownership structures are being fixed for the next 30 years. Whoever controls stakes in AI infrastructure, strategic data pools and compute capacity now holds a generational rent position. Among the top-7 hyperscalers, the outcome is largely decided. In the next wave — sector-specific AI, edge computing, robotics, biotech AI — it remains open. That is where sovereign funds and cooperatives must enter now, not in five years.
The strongest objection concerns timing. A sovereign wealth fund that buys only after valuations have expanded pays a multiple for the same stake and transfers the price risk to the public. Then comes the governance question: who decides how a fund of this size votes at general meetings? Norway needed decades of institution-building to address that, and the debate over exclusion criteria is still unresolved. Distributing ownership broadly also distributes power. The question of who then exercises it is not solved but deferred.
What we do not yet know
- How does a sovereign AI fund avoid becoming a vehicle for political capture (see Russia, Saudi Arabia)?
- What computing power must Europe control directly to avoid sliding into technological vassalage?
- How does one value "strategic" stakes that are not primarily financial within an investment mandate?
- What role do NGO and foundation structures (the Wikimedia Foundation model) play in AI commons?
- How does one prevent capital flight from jurisdictions that introduce strategic-stake mechanisms early?
Connected to: 01 · Taxes, 02 · Income, 06 · Democracy, 07 · Data & AI
Scenarios: Technology & AI · Economics & Capital · Investment Implications
