On the Continent

What's working in European tech. Curated daily.

AI / Foundation Models

High Dependency

Overview

Foundation models are overwhelmingly US-built (OpenAI, Google, Anthropic, Meta) or Chinese (DeepSeek, Alibaba). European companies and governments rely on these for AI capabilities. No European model currently operates at the frontier scale of GPT-4 or Claude.

Europe's strength lies in AI regulation (EU AI Act) and application-layer AI, but the foundational infrastructure — the models themselves — remains imported.

Why EU Sovereignty Matters

AI foundation models are becoming the operating system of the digital economy. If European businesses and governments depend entirely on foreign-built models, they cede control over a critical general-purpose technology. This affects everything from industrial automation to public services, scientific research, and defense applications.

Key European Players

Startup

Aleph Alpha DE series-b

German AI company focused on sovereign AI for European enterprise and government. Shifted from model training to AI infrastructure platform.

Mistral AI FR series-b

France's leading AI lab, building open-weight foundation models. Mistral Large and Mistral Small compete with GPT-class models. Raised over €1B.

Research

Swedish national AI research program that built GPT-SW3, a Nordic-language foundation model trained on European supercomputers.

Initiative

World's first comprehensive AI regulation. Risk-based framework that classifies AI systems and sets requirements for high-risk applications.

InvestAI EU

EU initiative targeting €200B mobilization for AI infrastructure, compute capacity, and European AI development.

Key Facts

Key policy
EU AI Act (2024), InvestAI initiative
Regulatory lead
First comprehensive AI regulation globally
Investment target
€200B (InvestAI mobilization)
EU frontier models
None at GPT-4/Claude scale

Recent News

Suggest an Edit

Know a European player we're missing? See something that needs updating? Suggestions are reviewed by our editorial team.

What would you like to suggest?