roobriefIT news for developers
Search

By AI

Mistral's €3 Billion Round: How Big Is Europe's AI Gap With the US?

Mistral’s €3 billion round shows Europe’s AI potential, but so far neither capital nor compute offers evidence that the gap with the US has narrowed.

French AI company Mistral announced on Sept. 8, 2026, that it had raised €3 billion (about $3.4 billion) in a Series D round. Its valuation after the round is above €21 billion (about $23.6 billion). Samsung Electronics led the round, with the Scaleup Europe Fund, managed by EQT, and existing investor PSG Equity as co-leads. It matters that industrial and public-sector capital has joined a large AI company from Europe.

Still, one funding round cannot stand in for Europe’s AI competitiveness. The gap has opened up not only because of model-building ability but because of three things working together: large-scale capital, computing capacity, and the ability to grow companies in a market split across many countries.

Why Europe Grew More Slowly Than the US

Start with the money. According to the 2026 AI Index from Stanford’s Institute for Human-Centered AI (Stanford HAI), private AI investment in 2025 was $285.9 billion in the US and $20.9 billion in Europe, roughly a 14-fold difference. The figures are a regional tally of AI investment, so “Europe” here should not be read as the EU. Even so, the pattern is clear: far more AI companies and follow-on funding gather in the US.

That money ties back to computing capacity. Bruegel, drawing on the Europe2031.ai dataset of 100 AI infrastructure projects of at least 10 megawatts, estimates that the EU had about 2 gigawatts of operational AI compute in 2026, roughly 5% of the world’s total. The US has about 35 gigawatts, or 78%. The tally leaves out capacity in existing data centers that large US cloud providers run inside the EU, so comparing the two numbers directly calls for caution. Still, without enough infrastructure to train and serve large models, European companies either depend on US cloud providers or, if they want to build their own, must secure huge amounts of capital and electricity.

Europe’s weakness does not mean it lacks research. A European Commission study of patent and publication data found that Europe’s research hubs are less connected to one another than US hubs, and that the gap is larger in complex fields such as AI. On top of that, the Draghi report on EU competitiveness diagnosed that the single market and an integrated capital market do not work well enough, so promising companies struggle to expand across countries or to raise large growth-stage funding. Turning ideas into businesses and building continent-scale companies is slow going.

Where Europe’s AI Strengths Lie Today

Ask who builds more cutting-edge models and who has secured the capital and compute behind them, and the US is ahead. The same AI Index shows Europe’s 2025 private investment far below the US, and Bruegel’s report also judges that the EU trails the US and China at every layer, from chips to models to applications. Mistral is counted among Europe’s few developers of cutting-edge models, but one company’s growth cannot close the gap for the whole ecosystem.

By country, France, with Mistral at its center, is the leading example of building its own models. When the Stanford HAI AI Index counted “notable AI models” released in 2025, France and the UK had one each. The count assigns models by the institutions of the paper authors, so it is not a ranking of national AI industries. On private investment, the UK, France and Germany led in Europe. In Germany, Fraunhofer institutes continue research on applying AI to production and logistics optimization. The UK is a European country but not an EU member, so it has to be kept apart from EU policy.

Switzerland, outside the EU, is hard to leave out. Apertus, released by ETH Zurich, EPFL and the Swiss National Supercomputing Centre, is a multilingual language model in two sizes, 8 billion and 70 billion parameters, with its training data and methods published as well. Switzerland’s result is less about private investment than about public research institutions building an open model and the training infrastructure together.

The foundation for applying AI in industry, and the research talent behind it, are another source of European strength. Sectors where established industry is strong, such as automotive, manufacturing, health care and science, are a different arena from general-purpose chatbots. In a McKinsey survey cited by the AI Index, the share of organizations saying they use AI in at least one business function was 91% in Europe and 90% in North America in 2025. Because it rests on survey responses, it does not mean Europe builds better models or captures more value, and it is a different measure from model development or investment size. It is safer to read it as showing no large gap in adoption.

Mistral, too, positions itself not as a head-on rival to US companies in consumer chatbots but as a provider of open-weight models, its own infrastructure and enterprise products together. Open-weight means releasing model weights so customers can download, run, or adapt them themselves. That can carve out a niche among European companies and public bodies that value control over their data and model operations, but open weights do not by themselves solve the problems of compute, deployment and winning customers.

Can EU Policy Close the Gap?

The EU is pursuing policies that build shared computing infrastructure and industrial adoption together. The AI Continent Action Plan, launched in 2025, aims to expand compute, data, talent and industrial uptake. Under it, the European Commission said in April 2026 that it had set up 19 AI Factories, public AI facilities built around supercomputers that help researchers and companies develop models, plus 13 regional access points.

An AI Gigafactory is a much larger computing facility for training, bigger than an AI Factory. In July 2026 the EU opened a tender to select up to seven of them, with up to €10 billion (about $11.2 billion) in EU and member-state public funding meant to draw at least €20 billion (about $22.5 billion) in private investment. Construction has not started yet. Turning it into real computing capacity takes more than choosing operators: investment has to be committed, power supplied, and data centers built and brought into service.

The Apply AI Strategy is a plan to spread AI use across 10 key industries, including small and medium-sized businesses, and in the public sector. In April 2026 the Commission said it had earmarked €1 billion (about $1.1 billion) in funding calls for industry and the public sector. The Data Union Strategy is meant to widen data access and sharing.

Calls to restrict AI continue as well. The AI Act, the EU’s main AI law, adopted in 2024, bans some uses such as social scoring and puts obligations like data quality, record-keeping and human oversight on high-risk areas. Part of the law began to be enforced in August 2026.

At the same time, the AI Omnibus, an amendment that simplifies the AI Act to lower the compliance burden, took effect on July 27. It extends support for small businesses and regulatory sandboxes and pushes some high-risk AI obligations back to 2027–2028. It also adds a ban on apps that generate non-consensual sexual images or child sexual abuse material, though that ban applies only from December 2026.

In January 2026, when the amendment was still a proposal, the European Data Protection Board and the European Data Protection Supervisor said in a joint opinion that they supported simplification but wanted stronger safeguards so that protection of fundamental rights is not weakened. We could not confirm how far the final text reflects those demands. The debate now centers less on a blanket ban on AI than on how far to block risky uses and how to reduce the burden on companies.

What to watch

In the end, Mistral’s funding round is less a declaration that the gap has closed than a case study in what Europe has to build to stay in the race. The things to watch are whether the EU’s AI Gigafactories break ground and come online, follow-on investment in growth-stage AI companies, and the results of AI adoption in manufacturing and the public sector. If all three follow, Europe’s industrial base and research strength could turn into competitiveness.

Dollar figures for euro amounts use the Oct. 21, 2026 exchange rate (1 euro ≈ $1.12) and are approximate.

Sources

Source: mistral.ai

← Back