Advanced Materials Are Becoming a Foundation for AI Infrastructure

Summary

This sponsored MIT Technology Review Insights article argues that the expansion of AI is becoming a materials challenge as semiconductors and data centers approach limits involving heat, electrical efficiency, purity, chemical resistance, plasma resistance, and reliability. Syensqo executive Mike Finelli describes work on materials for high-voltage data-center architectures, semiconductor-fab seals, direct-immersion cooling, and other thermal-management systems. The company is also transferring materials knowledge from electric-vehicle batteries and power systems to data centers, including insulating materials, battery binders, and cooling technologies. Finelli says sustainability is now part of the definition of performance: Syensqo assesses research projects at the outset and reports that 88% of its portfolio qualifies as sustainable under its internal framework. The article also describes a roughly two-year effort with Microsoft using AI agents to digitally synthesize millions of molecular combinations, predict physical, chemical, toxicity, and sustainability characteristics, rank candidates, and reduce the laboratory shortlist to about 100 materials. Finelli emphasizes that these systems support rather than replace scientists by reducing search time. He envisions a feedback loop in which AI helps create materials that improve AI infrastructure, enabling further advances in AI-assisted materials discovery. The claims are presented in a sponsored interview with Syensqo and should be understood as the company’s account of its work.