Altrove: AI-Powered Material Discovery to Address Rare Earth Geopolitical Vulnerability
Clip title: AI Meets Materials | Altrove - Joonatan Laulainen Author / channel: The Bottomline Podcast URL: https://www.youtube.com/watch?v=MITjX5LLRAc
Summary
The video features Joonatan Laulainen, Co-founder and CTO of Altrove, a Paris-based startup leveraging AI and lab automation to design alternatives to critical materials like rare earth magnets. The discussion centers on the geopolitical vulnerabilities arising from the world’s heavy reliance on a single country, China, which controls approximately 90% of rare earth refining capacity. Despite past incidents, such as China slashing export quotas in 2010, the issue was not treated as a crisis due to a global focus on cost-efficiency and a perceived cooling in international geopolitics. The demand for these materials has surged with the rise of electric vehicles, wind turbines, and advanced electronics, intensifying the urgency to find independent and sustainable alternatives.
Altrove’s approach to material discovery involves a “full-stack” methodology, integrating computational prediction with experimental validation. They reverse-engineer requirements from industrial partners, focusing on specific material properties, manufacturing processes, cost, and supply chain constraints. This collaborative process begins by defining a “spec sheet” outlining the desired material characteristics. Altrove then utilizes a combination of machine learning models, AI algorithms, simulations, and expert judgment to generate and narrow down potential novel material candidates. A crucial step in their internal process is to quickly assess the “synthesizability” of these proposed materials, prioritizing speed in development to drastically reduce the typical 20+ year lab-to-market timeline to 15-18 months.
The conversation further delves into the broader challenges and future trajectory of AI in materials science. Joonatan draws parallels with biotech, noting the immense “search space” for new materials and the need for robust databases and generative models. A key distinction is made between “rediscovery” (predicting known materials) and true “discovery” (novel materials), with many AI models currently excelling more at the former due to training on existing data. Altrove addresses this by emphasizing the generation of high-quality, proprietary experimental data in their labs, which is crucial for training AI models that can accurately predict and optimize real-world material properties. This integrated computational and experimental capability allows them to both identify promising candidates and rapidly produce tangible samples for industrial partners, directly addressing market needs and driving innovation in a highly R&D-intensive sector. The success in this field, he concludes, hinges on the ability to marry smart hypothesis generation with efficient, data-driven experimentation and a clear pathway to scalable production.
Video Description & Links
Description
Every material the modern economy depends on eventually has to leave the lab and enter a factory, and that transition is where most materials science breakthroughs quietly die. A model can predict a compound in seconds; manufacturing it at scale has historically taken decades. That gap is not abstract. Over 90% of the world’s rare earth magnet refining sits inside one country, and when export controls tightened in 2025, the fragility of that dependency became impossible to ignore. The real question was never just who could discover a better material. It was always who could actually build one before the world ran out of time to wait.
In this episode of Bottomline, Drishti sits down with Joonatan Laulainen, Co-Founder and CTO of Altrove, a Paris-based deep-tech startup building AI-driven, lab-integrated infrastructure to design and manufacture alternatives to critical materials like rare earth magnets. Together, they trace how China’s refining dominance was decades in the making, why Altrove chose to own physical synthesis instead of stopping at computational prediction, where the real intellectual property sits across a value chain most people assume is flat, why the field keeps mistaking rediscovery for genuine discovery, and what it takes to compress a process that has historically taken decades into months.
ABOUT ALTROVE
Altrove is a Paris-based deep-tech startup building AI-driven, lab-integrated infrastructure to discover and manufacture alternatives to critical materials such as rare earth magnets. Where most AI-for-materials companies stop at computational prediction, Altrove closes the loop with an in-house automated lab, synthesizing and testing every candidate to build a proprietary data advantage competitors can’t replicate from public data alone. The company partners directly with industrial clients on a licensed basis, converting material specifications into synthesizable, validated solutions engineered to scale from day one.
Key Topics China’s rare earth monopoly, and the expertise the West let slip away Altrove’s full-stack bet, from AI generation to physical synthesis The AI materials value chain, and where the real IP actually sits Rediscovery versus discovery, and AI’s bias toward the plausible Compressing materials timelines, and the bet on scaling speed over efficiency.
Related Concepts
- rare earth magnets
- material discovery
- AI-driven design
- lab automation — Wikipedia
- geopolitical vulnerability
- critical materials — Wikipedia
- supply chain resilience — Wikipedia
- AI-driven material discovery
- generative models — Wikipedia
- export controls — Wikipedia
Related Entities
- Altrove — Wikipedia
- Joonatan Laulainen
- China — Wikipedia