The Rules

Governance, Regulation & Sovereign AI

Cross-cutting force · The rules the city runs under

Why it matters
Rules determine which architectures are permissible, which suppliers are usable in a given jurisdiction, and how much documentation sits between a model and production.
The bottleneck
Fragmentation. Different jurisdictions are converging slowly, and multinational deployments have to satisfy the strictest applicable rule.
Who captures value?
Providers who make compliance straightforward, including sovereign infrastructure operators and governance tooling, plus incumbents whose scale absorbs compliance cost more easily.
What could change?
Enforcement intensity is still being established. Meaningful penalties under existing frameworks would change buying behaviour faster than any new statute.

Rules shape what can be built, where data may sit, and what a deployed model must be able to explain about itself. They cut across every layer rather than sitting on top of one.

Regulation is often treated as an afterthought in technology writing, which is a mistake here. Data residency requirements decide which data centres a customer may use. Model governance obligations decide what documentation a lab must produce. Export controls decide who may buy the most advanced chips. These are not commentary on the ecosystem, they are inputs to it.

THE RULES THE CITY RUNS UNDERExport controlsGovernments decide who may buy the most advanced chips and the machines that make them. Trade policy is now part of the supply chain.Data residencySome data may not leave a jurisdiction, which decides which data centres and clouds a customer can use at all.Model governanceThe EU AI Act sets obligations by risk category. Documentation, testing and disclosure requirements sit between a model and production.Privacy lawGDPRAI systems processing personal data sit under the same privacy law as everything else. Training data and inference logs both count.Sovereign AIGovernments funding domestic compute and models so that strategic capability does not depend entirely on foreign suppliers.Rules are inputs to the ecosystem, not commentary on it.

Tap or hover a box to see what it does

Rules decide what can be built, where data may sit and what a model must explain. They are inputs to the ecosystem, not commentary on it.

Regulation and model governance

The EU AI Act sets obligations by risk category, and frameworks such as the NIST AI Risk Management Framework provide voluntary structure. Privacy law, notably GDPR, applies to AI systems as it does to anything else processing personal data.

Sovereignty and national infrastructure

Several governments now treat domestic compute and models as strategic assets, funding national infrastructure and requiring that certain data and workloads remain inside their borders. That is a large part of why regional AI clouds exist.

publicNBIS

Nebius Group

What it does

Operates AI focused cloud infrastructure in Europe. It emerged from the old Yandex international business after the Russian operations were sold in 2024, with the Dutch parent keeping the global assets.

Why it matters

European customers with data residency requirements need somewhere local to train and serve. Sovereign preference is a real commercial tailwind.

What could go wrong

Sub-scale relative to the hyperscalers, capital hungry, and still proving durable demand.

Position in the ecosystem
private

Mistral AI

What it does

Develops efficient models, several released with open weights, aimed at European enterprises and sovereign deployments.

Why it matters

The clearest European answer to the question of who trains models for customers that would rather not send data across the Atlantic.

What could go wrong

Competing on efficiency against far larger budgets, and a sovereignty tailwind that is partly political.

Public market exposure

Private, French, with strategic investors including industrial and technology partners. Exposure is largely unavailable to public investors.

Position in the ecosystem
publicORCL

Oracle

What it does

Runs Oracle Cloud Infrastructure with a strong GPU cluster business, and sells the databases and applications that hold a great deal of enterprise data.

Why it matters

It sits in two layers at once. The database estate gives it a claim on enterprise data, and OCI has become a meaningful venue for large scale AI training capacity.

What could go wrong

Building GPU capacity is capital hungry and the customer list is concentrated. Committed backlog is a promise, not yet revenue.

Position in the ecosystem

How to think about the economics

  • GrowthModerate

    How quickly demand in this part of the ecosystem is expanding.

  • Capital intensityLow

    How much money has to be spent up front before revenue arrives.

  • Competitive moatModerate

    How difficult it is for a credible new entrant to take the business.

  • Customer concentrationLow

    How much revenue depends on a small number of buyers.

  • Disruption riskModerate

    How exposed the layer is to a technical or commercial shift.

This is a framework for thinking about the economics of a layer, not a recommendation. An important AI company, a strategically advantaged company, an investable security and an attractively valued security are four different things.

Where value may accrue

Trust, jurisdictional presence and the ability to evidence compliance. Compliance cost also acts as a quiet barrier to entry, which tends to favour scale.

Key terms in this layer

Sovereign AI
Compute, models and data kept inside a jurisdiction because governments treat them as strategic assets. A large part of why regional AI clouds exist.
ExampleNational AI infrastructure programmes in Europe and the Middle East.
Related termsData centreFoundation model
Sources and further reading (3)+
  1. 1Regulation (EU) 2024/1689, the AI Act · European Union
  2. 2AI Risk Management Framework · NIST
  3. 3General Data Protection Regulation · European Union
Last fact-checked: 29 August 2026

The AI ecosystem changes rapidly. Company positions, technologies and market data reflect information available at the date above.