Lithosquare raises $25M seed to apply foundational AI models to critical metals exploration

Lithosquare raises $25M seed to apply foundational AI models to critical metals exploration

Paris-based Lithosquare has raised a $25M equity seed round co-led by World Fund and Kindred Capital, with participation from Daphni, Omnes Capital and Ovni Capital. The startup applies a multimodal foundational AI model—trained on geological maps, textual reports, geophysical data (magnetism, radioactivity), and surveys—to accelerate the discovery of critical metals deposits such as copper, lithium, and nickel. The model differs from traditional pattern-recognition tools by encoding geological science rather than just matching known deposit signatures, enabling exploration in new geographies and deeper, previously unprospected sites. Lithosquare announced its first public partnership with UK mining company Aterian, covering eight exploration projects in Morocco and Botswana, and typically monetizes via upfront baseline fees plus future outcome-based payments and revenue-sharing on potential mines.

Why it matters: This fundraise signals the early institutionalization of a model-centric approach to natural resource discovery—an emerging vertical application of foundation models that fits the "domain-specific fine-tuning" frame within the broader AI substrate. Lithosquare’s approach updates the geological exploration playbook by moving beyond pattern-recognition machine learning to a generative/hypothesis-testing paradigm, reducing analysis timelines from months to days. The capital comes at a modest scale compared to U.S. rival KoBold Metals ($537M Series C at $3B valuation), but positions Lithosquare as the leading European contender in a segment increasingly tied to sovereignty imperatives, given the EU's Critical Raw Materials Act and geopolitical pressure to diversify supply chains. The round is notable less for its size—well below the $500M cross-subsection capital threshold—and more for what it validates: that foundation-model reasoning can be productized for physical-world exploration, a thesis that sits at the intersection of segment 10 (robotics/physical AI) and the broader geoscience data-infrastructure layer.

Grounded expert take: Lithosquare’s differentiation lies in its foundational model design—trained on multimodal geological data rather than supervised on known deposits—which could resolve a key limitation of prior exploration AI: the inability to discover deposit types that don’t resemble existing ones. If the technology delivers on its promise of opening new geographies and deposit classes, it could materially alter the cost structure and timeline of critical-minerals supply chains, a structural force with cross-substrate implications for the energy transition and compute hardware manufacturing. However, the startup faces two canonical risks: proving that its model generalizes across geologies at scale, and competing with well-capitalized U.S. peers like KoBold Metals that have deeper reserves of both data and capital. The $25M seed provides runway to double the team to 40 and open a North America office—but the real signal to watch is whether Lithosquare secures a second major partnership or a larger institutional round before the model's outputs translate into confirmed economic discoveries.

#AIforDiscovery #CriticalMinerals #FoundationModels #EnergyTransition #GeospatialAI #EuropeanDeeptech