Saint-Gobain says Microsoft AI helped generate cement-additive ideas in six weeks
The research drew on 62,000 patents and 44,000 abstracts. Candidate molecules still need to prove they can help reduce cement’s carbon footprint.
Loading page…
The research drew on 62,000 patents and 44,000 abstracts. Candidate molecules still need to prove they can help reduce cement’s carbon footprint.
Listen to this story
Saint-Gobain is using Microsoft Discovery to find additives that could reduce cement’s clinker requirement beyond the roughly 20% reduction achieved by its existing EnviroAdd product, while also strengthening cement. The company says researchers developed several candidate-molecule hypotheses in about six weeks, but the candidates are still being tested and no performance results are established. Program manager Mojdeh Rastgoo-Lemaitre says AI cut the time to build the research knowledge base from an estimated year to a couple of hours; that comparison comes from Saint-Gobain, not an independent assessment.
The team searched 62,000 patents and 44,000 scientific abstracts, with cross-disciplinary literature surfacing molecules researchers might otherwise have missed.
If candidates pass further scrutiny, a development team will test them under physical conditions.
Saint-Gobain joined Discovery’s private release in March 2026, and the companies signed a sustainable-materials framework agreement in July.
Saint-Gobain says its researchers reached new ideas for lower-carbon cement additives in about six weeks using Microsoft’s AI research platform. The candidate molecules are still being tested. Microsoft published the customer account on October 20, 2026; the reported speed gains come from Saint-Gobain, not an independent assessment.
The team gathered 62,000 patents and 44,000 scientific abstracts using Microsoft Discovery. Researchers then narrowed and analyzed that material to develop several hypotheses about ingredients worth testing.
Discovery lets scientists direct AI agents—software systems assigned research tasks—to search literature, connect information, generate hypotheses, design experiments and run analyses. For this project, Saint-Gobain program manager Mojdeh Rastgoo-Lemaitre said searching beyond construction research surfaced molecules the scientists would not otherwise have considered.
Rastgoo-Lemaitre’s largest time-saving estimate concerned creating the knowledge base, rather than the full process of developing an additive:
Without using this technology, it would have taken our researchers a year to create the knowledge base. Using AI, it took us a couple of hours.
Mojdeh Rastgoo-Lemaitre, digital acceleration program manager at Saint-Gobain
Clinker is the binding ingredient in cement. It is made by heating limestone to very high temperatures, a major source of cement production’s carbon emissions. Saint-Gobain wants a new additive that would reduce the amount of clinker needed while also making the cement stronger.
The company says its existing cement additive EnviroAdd already reduces the need for clinker by about 20%. The new research seeks to push that reduction further. If the candidates withstand scrutiny, a potential solution would move to a development team for testing under physical conditions.
Saint-Gobain was among Discovery’s first private-release customers in March 2026. Teams from both companies worked on fine-tuning the platform. Microsoft’s participating engineers came from its Frontier Company business, which places engineers inside customer organizations to deliver AI solutions.
The companies signed a framework agreement in July to accelerate work on sustainable construction materials and processes. In an October 20 Microsoft France announcement, Microsoft said Saint-Gobain was continuing to extend its use of Discovery following that agreement.
Loading discussion...
Join the conversation
Explain whether saving researchers time is enough, or results should come first.
Be the first to share a perspective or an experience.
Reader comments
Newest comments first. Replies stay oldest first.