The money is no longer flowing into artificial intelligence. It is flooding in.
From Paris to San Francisco, AI companies are raising sums that would have looked absurd even eighteen months ago. Venture capital, once spread across fintech, climate, biotech and consumer apps, is now concentrating around a handful of frontier AI firms, data-centre builders and chip suppliers.
In Europe, the reaction is increasingly uneasy: admiration mixed with alarm that the continent may once again arrive late to a technological platform shift.
Venture Capital Collapses Into AI
According to figures referenced by Crunchbase and investors tracking the sector, global startup funding reached roughly $300 billion in the first quarter of 2026, with around 80% tied directly to AI-related companies and infrastructure.
Four deals alone reportedly represented nearly two-thirds of all venture funding during the quarter.
The largest beneficiary remains OpenAI.
The company closed a historic funding round this spring worth more than $120 billion, pushing its valuation above $850 billion. An AI lab founded less than a decade ago is now valued higher than many European stock exchanges combined.
But the more startling numbers may belong to Anthropic.
The company reportedly told investors it expects second-quarter 2026 revenue of $10.9 billion and an operating profit of roughly $559 million — potentially its first profitable quarter.
Three years ago, most AI executives openly admitted they had no realistic path to profitability because computing costs were devouring revenue as fast as it arrived. That assumption is beginning to crack.
AI Labs Are Becoming Infrastructure Giants
Profitability has not slowed spending. It has accelerated it.
OpenAI is projecting hundreds of billions in future infrastructure expenditure. Anthropic has signed long-term compute agreements with cloud and infrastructure providers that could eventually total hundreds of billions of dollars.
AI firms are no longer behaving like software startups.
They are behaving like utilities.
The real race is no longer only about models. It is about:
- electricity supply
- land acquisition
- cooling systems
- data-centre construction
- access to advanced chips
Stanford University’s 2026 AI Index estimates annualised revenues for leading AI firms at extraordinary levels:
| Company | Estimated Annualised Revenue |
|---|---|
| OpenAI | ~$25 billion |
| Anthropic | ~$19 billion |
| xAI | Rapidly rising |
| Mistral AI | Smaller but expanding aggressively |
At the same time, compute expenditure continues climbing alongside revenue growth.
Europe Tries to Build Its Own AI Backbone
In Europe, the anxiety has become public.
Mistral AI — often described as Europe’s strongest answer to American AI dominance — has spent the past year transforming itself from a model developer into an infrastructure company.
In March, the company secured roughly $830 million to help build new data-centre capacity near Paris.
Earlier this year, it announced a €1.2 billion investment connected to Swedish AI infrastructure projects.
Last week, Mistral acquired Austrian startup Emmi AI, a specialist in industrial physics simulations used in aerospace, semiconductor and manufacturing systems.
The acquisition was relatively small by Silicon Valley standards, but politically symbolic.
Europe increasingly sees industrial AI — not chatbots — as its most realistic strategic opening.
“Two Years” to Avoid Dependency
During a hearing before the French National Assembly this month, Mistral chief executive Arthur Mensch warned that Europe had roughly “two years” to avoid technological dependency on foreign AI infrastructure.
The warning reflected a growing fear inside Brussels and Paris:
Europe risks losing control not only over software platforms, but over the physical computing backbone powering the next economy.
That concern is now shaping industrial policy.
A French consortium backed by companies including:
- Capgemini
- Orange
- EDF
- Scaleway
is preparing a bid for support from the European Union’s new €20 billion AI infrastructure initiative.
The proposal reportedly includes a €10 billion French data-centre project designed to expand Europe’s sovereign compute capacity.
Regulation Is Not Infrastructure
Behind the headlines sits a more uncomfortable reality.
Much of Europe’s AI ambition still depends on:
- foreign chips
- foreign cloud providers
- foreign capital
Even Mistral, Europe’s flagship AI startup, relies heavily on partnerships and financing structures connected to non-European technology ecosystems.
Meanwhile, investors continue rewarding scale above all else.
The venture market is increasingly concentrating around a tiny number of firms capable of building frontier models and securing massive compute access.
Smaller startups are becoming suppliers, application layers or acquisition targets.
The old venture-capital logic of spreading risk across hundreds of companies is giving way to infrastructure concentration around a few dominant AI platforms.
Europe’s AI Debate Has Shifted
For years, Europe approached artificial intelligence primarily through regulation:
- privacy rules
- competition law
- the AI Act
But regulation does not create GPU clusters, power grids or hyperscale computing facilities.
Investors now measure AI competitiveness in gigawatts as much as algorithms.
And while European officials continue discussing digital sovereignty, the market is already moving at infrastructure speed.
The industry increasingly resembles railroads, telecoms and energy more than traditional software.
Capital expenditure is ballooning. Margins remain uncertain. Valuations continue climbing anyway.
For now, investors appear willing to believe that whoever controls compute will eventually control everything built on top of it.
Subscribe to EuroLuminant for independent European journalism.
