Susanna Potenza, Investment Associate at Primo Capital, shares her perspective on AI4Science and climate transition—key issues for Primo Climate, Primo Capital's fund that invests in decarbonization technologies.
While pharmaceutical companies active in drug discovery have historically been accustomed to investing heavily in the search for new molecules, the same has not been true—at least until now—for large industrial companies in the search for new materials. However, the discovery of innovative materials is taking on increasingly central importance in light of the challenges associated with decarbonization.
Decarbonization today means the massive, global, and repeated adoption of already validated technologies, such as photovoltaics and lithium batteries.
The risk, however, is that once the problem of operational CO₂ emissions has been solved, we will fall back into a familiar cycle: that of dependence on a few critical elements, extracted and exploited at rates that potentially exceed the Earth's capacity to regenerate its resources.
For this reason, at Primo Climate we have analyzed the topic of AI for Science, i.e., the possibility of conducting research on materials by combining atoms in innovative ways thanks to artificial intelligence (AI). AI allows us to exploit computing power, input data quantities, and simulation scales that are infinitely superior to those of a single research center. The goal is to identify new materials with mechanical, chemical, and physical properties currently provided by “unique” materials, such as neodymium in magnets. The supply chain for the latter already presents—or will present in the future—significant challenges (think, for example, of neodymium extraction or the potential over-exploitation of lithium).
In this sense, the evolution of climate tech must move along two complementary lines:
on the one hand, transforming waste into new mines, pushing towards ever deeper and more systemic circularity;
on the other, innovating upstream, identifying new molecules and new materials capable of offering performance similar to or superior to current ones, without the same environmental and geopolitical criticalities.
It is in this context that materials research can now be enhanced by AI, which represents a substantial acceleration of the discovery process thanks to its extensive computational capacity to test, simulate, and iterate rapidly.
Both major players and highly innovative startups are entering this space, where scientists and engineers are working on different replacement “recipes” with diverse industrial applications.
Artificial intelligence is now used throughout the entire materials discovery chain, starting with the selection of promising molecules from tens of millions of possible structures. Algorithms learn molecular representations that allow the structure and properties of materials to be correlated. The main operating modes include:
Atomistic simulations (e.g., MLIP and Graph Neural Networks models) to predict stability and performance;
Validation with quantum methods (DFT) to refine candidates;
Physical testing in the laboratory, which feeds back into the models in a continuous cycle of active learning.
The main applications explored so far, both by large companies and startups, focus on key sectors for the energy transition: batteries (new electrolytes and electrodes alternative to lithium or vanadium), fuel cells (non-noble catalysts to replace platinum and palladium), semiconductors (materials with greater conductivity and thermal efficiency), green hydrogen and ammonia, through more efficient and durable electrodes and electrolytes.
Today's ecosystem is populated by incumbents (major technology players) as well as emerging startups. The big players—such as Google DeepMind, MIT, IBM, BASF, and Siemens—are developing AI platforms and automated laboratories, but often with a strong academic focus or limitations on commercial use. At the same time, deep-tech startups such as Periodic Labs, Lila Science, Orbital Materials, Dunia, and Entalpic are combining proprietary AI and self-driving labs to rapidly innovate the new materials sector.
However, there is a substantial difference: in the United States, the sector has already raised hundreds of millions of dollars (the total raised by the main players exceeds $900 million), while Europe lags behind in terms of funding. From a geographical point of view, the Fund's main target is the Italian market (at least 80% of the commitment), without however excluding the most interesting opportunities in Europe (including Great Britain, Switzerland, and the Republic of San Marino). The IEA's Net Zero report underscores the pivotal role of innovation, revealing that in a net-zero emission scenario, 48% of the reduction in CO2 emissions will originate from technologies that are currently in the demonstration or prototype phase.
.png)
We therefore believe that a strategic scenario is opening up in which materials research, in order to become competitive in Europe and above all to avoid new concentrations in the supply chain, will necessarily have to take a step up, passing through a few rare and highly innovative startups capable of integrating advanced scientific research and artificial intelligence.
