Europa Nova

The question is not whether artificial general intelligence will arrive. The question is whether Europe will have any say in what it becomes.

The Uncomfortable Truth

We address this letter to you not with alarm, but with clarity. As you read this, the most consequential technology in human history is being developed almost exclusively by a handful of private American and Chinese corporations. The decisions these organisations make about artificial general intelligence — how it reasons, what values it embodies, whose interests it serves — will reshape every economy, every institution, and every society on Earth within the next decade.

Europe has no seat at this table. Not because we lack the talent, the scientific tradition, or the resources. But because we have failed to commit. We have authored regulations while others built infrastructure. We have published strategies while others trained models. We have distributed funding across 27 nations in the name of fairness while our competitors concentrated resources in the name of results.

This letter proposes a different path. Not another framework, not another action plan, not another consortium of committees. We propose a single, decisive act of sovereign ambition: the creation of Europa Nova — an independent, treaty-based European institution dedicated to the development of artificial general intelligence that reflects European values, serves European citizens, and positions Europe as a co-author of humanity’s most powerful technology.

Why This Moment Matters

Three developments have converged to create a narrow but real window of opportunity for Europe.

1. The Architecture Is Not Settled

The Transformer architecture that underpins today’s leading models is almost certainly not the endpoint. Research into state-space models, continuous-time architectures, and hybrid attention mechanisms suggests that fundamentally more compute-efficient approaches are within reach. A European initiative that begins now is not entering a race already lost — it is positioning itself for the next paradigm, where architectural innovation matters more than brute-force scaling.

2. Energy Has Become the Decisive Constraint

Training frontier AI models requires extraordinary amounts of electricity. The established data centre hubs in Frankfurt, London, Amsterdam, and Dublin are facing severe grid congestion. Wait times for grid connections in EU countries now range from two to ten years. Meanwhile, the Nordic countries — Norway, Sweden, Finland, and Iceland — possess an abundance of renewable hydroelectric and wind power at a fraction of the cost. Northern Europe’s cool climate reduces cooling costs by 30–60% compared to conventional locations. Europe’s perceived weakness — the absence of concentrated tech infrastructure — becomes a strength when the bottleneck shifts from software to energy.

3. The Geopolitical Window Is Open

Current US policy on AI export controls, combined with growing concerns about digital sovereignty, means that dependence on American AI infrastructure carries real strategic risk. The EU has recognised this through its AI Continent Action Plan and InvestAI initiative. But recognition is not the same as execution. Europe needs a vehicle that can move at the speed of technology, not the speed of bureaucracy.

What We Propose: Europa Nova

Europa Nova is a sovereign European research institution, established by international treaty, dedicated to the development of safe artificial general intelligence. It is modelled not on EU agencies, but on CERN — an institution that proved Europe can lead the world in fundamental research when it commits fully.

But Europa Nova differs from CERN in one critical respect: AGI is not only a scientific endeavour. It is an economic and strategic one. Therefore, Europa Nova is designed from the outset to translate research into industrial capability, with a commercial licensing arm that ensures Europe captures the economic value of its investment.

Organisational Structure

The Nordic Compute Advantage

The location of Europa Nova’s compute infrastructure is not a political choice. It is an economic and physical necessity.

The market has already validated this logic. In July 2025, OpenAI announced Stargate Norway, its first European AI data centre, located near Narvik in northern Norway. Microsoft, CoreWeave, and Brookfield collectively committed over $15 billion to Nordic AI infrastructure in 2025 alone.

Funding: Ambitious but Realistic

Europa Nova requires a committed investment of €50 billion over ten years, with a minimum of €5 billion annually. This is substantial. It is also modest compared to what the competition is spending. OpenAI’s Stargate initiative alone represents $500 billion in planned US AI infrastructure investment.

Treaty-based sovereign contributions from 5–8 participating nations provide the foundational €3 billion per year. This is not subject to annual budget negotiations. It is a treaty obligation.

Strategic sovereign wealth fund allocation. Norway’s Government Pension Fund Global, the world’s largest sovereign wealth fund at approximately $1.7 trillion, represents a unique European asset. A dedicated allocation of 1–2% — structured as an investment, not a grant — would provide anchor capital.

Commercial revenue from licensing is projected to contribute meaningfully from year 4–5 onward, progressively reducing the need for sovereign contributions.

European Investment Bank co-financing for physical infrastructure leverages the EIB’s existing commitment to AI gigafactory financing.

Governance: Designed to Succeed, Not to Compromise

The single greatest risk to Europa Nova is not technical. It is organisational. European large-scale technology projects have repeatedly failed because governance was designed for political legitimacy rather than operational excellence. Europa Nova is designed differently.

A small founding coalition, not a continental committee. Five to eight nations with demonstrated AI capability and political will: France, Germany, the United Kingdom, Norway, Sweden, the Netherlands, Switzerland, and Finland.

Scientific leadership with real authority. The Scientific Director is recruited from the global top tier of AI research and has veto power on all technical decisions. Compensation is set at market rates — competitive with Anthropic, Google DeepMind, and OpenAI.

A rigorous sunset mechanism. After three years, an independent external panel conducts a comprehensive evaluation. If Europa Nova is not on a credible trajectory, the treaty includes provisions for orderly wind-down. This is not a concession to sceptics. It is a commitment to excellence.

No geographical distribution of research. One primary research hub. One primary compute facility. Not a “centre of excellence” in every participating nation. CERN works because it is in Geneva. Europa Nova works because it is concentrated.

The Talent Equation

Europe produces world-class AI researchers and then loses them. Yann LeCun is French — he works at Meta in New York. Demis Hassabis is British — DeepMind is owned by Google. This brain drain is not inevitable. It is the consequence of a gap that Europa Nova is specifically designed to close.

Three factors attract and retain top research talent: compensation, research freedom, and compute access. Europa Nova addresses all three. A fourth factor is often overlooked: quality of life. Zurich, Amsterdam, Stockholm, and Oslo are consistently ranked among the world’s most liveable cities — a genuine competitive advantage over San Francisco.

A European Approach to Values and Openness

Stratified openness. Fundamental research and architectural innovations are published openly. Pre-training models up to a defined capability threshold are released with open weights. Models above the threshold are available through controlled licensing. Alignment and safety research is always open.

Annotated diversity, not artificial balance. Europa Nova’s training data does not attempt to artificially balance perspectives. Instead, data sources are annotated with provenance, perspective, and context as metadata. The model learns that perspectives exist and differ — it does not learn that all perspectives are equally prevalent.

Democratic oversight with technical competence. An Advisory Council of parliamentarians provides democratic accountability, but without operational control.

Proposed Timeline

A Call for Sovereign Ambition

We have no illusion about the difficulty of what we propose. It requires political courage to concentrate resources rather than distribute them. It requires institutional innovation to create a structure that moves at the speed of technology rather than the speed of diplomacy. It requires financial commitment at a scale that will face scrutiny, and a tolerance for the possibility of failure that is foreign to most public-sector thinking.

But consider the alternative. Without sovereign AGI capability, Europe becomes a consumer of intelligence designed elsewhere, embedded with values chosen by others, governed by rules written without European input. Every European industry — automotive, financial services, healthcare, energy, defence — will depend on AI systems over which Europe has no influence and no recourse.

Europa Nova is not an expense. It is the investment that determines whether Europe remains a subject of history or becomes its object.

We invite the heads of state of European nations — beginning with those who recognise both the urgency and the opportunity — to convene a founding summit within the next twelve months. Not to discuss whether Europe should act, but to decide how.

The technology will not wait. The question is whether we will.