Among the weak signals that we like to decode, that of the investment portfolio of entrepreneurs occupies a special place. If the startups that make it up are not always successful, they reveal underlying beliefs, ecosystems and networks that their initiators seek to structure. We invite you to open this series with one of the most visible figures of artificial intelligence in Europe, Arthur Mensch.
As artificial intelligence establishes itself as a strategic infrastructure, the battle now lies in its implementation, its connection to data and its integration into operations. It is precisely in this area that Arthur Mensch, co-founder of Mistral AI, deploys his personal investments, with a portfolio that is still limited in number of participations, but which outlines the layers essential to the industrialization of AI and sheds light on the way in which a new generation of entrepreneurs is already considering the post-model.
Measure, before automate: BRAINTRUST
With Braintrust (US), the topic is not content generation, but its reliability. The startup develops tools to measure, test and improve the performance of AI systems in production.
With approximately 103 million euros raisedBraintrust positions itself in the model management and evaluation segment.
Connecting AI to reality: LINKUP
Another point of friction: access to information. Models, no matter how good they are, remain limited without connection to up-to-date data.
Linkup (France) develops an infrastructure allowing agents and LLMs to access the web and premium sources. The company has raised approximately 11.2 million eurosincluding a seed round in February 2026.
Automate critical functions: PARALLEL
In health, AI is now attacking the very organization of establishments.
Parallel (France) develops agents capable of handling administrative tasks (medical coding, billing, records management) by integrating with existing systems. The startup has just raised approximately 20 million eurosincluding a Series A round.
Same logic at Biorce (Spain), which intervenes upstream, at the level of clinical trials. With 52.3 million euros raisedof which 43.8 million euros in series Athe company targets a segment where lead times and costs remain structurally high.
Standardize agent production: LEMNI
As use cases multiply, how can we industrialize the creation of agents? This is the question that Lemni answers, by simplifying their design and deployment. The startup, financed to the tune of approximately 3 million eurosis aimed directly at companies wishing to integrate AI without mobilizing significant technical resources.
Specializing in generative uses: MIRELO AI
Finally, with Mirelo AI (France), the scope of application widens to the production of audio content. The company develops tools to automatically generate sound effects and musical compositions. She raised almost 34.9 million eurosnotably with Sequoia.
A value chain under construction
Taken individually, these investments cover different sectors. Together, they draw a coherent chain:
- connecting AI to data (Linkup)
- automate tasks (Parallel, Lemni)
- optimize critical verticals (Biorce, Mirelo)
- measure and manage systems (Braintrust)
