Advanced Materials Emerge as Critical Bottleneck for AI Infrastructure

The race to build more powerful AI systems is running into an unexpected constraint: the physical properties of the materials that underpin computing infrastructure. According to Mike Finelli, chief technology and innovation officer at Syensqo, AI workloads are pushing semiconductors and data centers toward performance thresholds that demand fundamentally new material solutions.

"AI is now, from a material standpoint, really pushing semiconductors and the data centers to their physical limits," Finelli told MIT Technology Review's Business Lab podcast. The challenge goes beyond simply supporting AI innovation—advanced materials are "actually increasingly defining what's going to be possible."

The 'and, and, and' principle

Finelli describes the materials challenge using what he calls the "performance pyramid." Commodity materials sit at the base, while high-performance specialty materials occupy the top. AI applications require materials that simultaneously meet multiple demanding criteria: high temperature tolerance, extreme purity, electrical performance, chemical resistance, plasma resistance, and long-term stability.

Each additional requirement—what Finelli calls the "and, and, and principle"—pushes materials toward the pyramid's apex. As AI advances accelerate, the number of simultaneous requirements multiplies, creating unprecedented demands on materials science.

Cross-industry solutions

Syensqo is developing materials across several fronts, including high-voltage architectures for next-generation data centers, advanced sealing materials for semiconductor manufacturing equipment, and thermal management solutions such as fluids for direct immersion cooling.

Some innovations transfer between industries. Materials originally developed for electric vehicle batteries and thermal management systems are now addressing similar challenges in data centers, which face higher voltage and energy density requirements. Battery energy storage systems designed for automotive applications are being adapted to smooth peak loads and provide backup power for data center operations.

Sustainability without trade-offs

Performance requirements now extend beyond technical specifications. Finelli notes that customers increasingly expect materials to meet sustainability targets alongside traditional performance metrics. "Our goal is to remove the trade-off between performance and sustainability," he said.

Syensqo addresses this by integrating sustainability considerations at the research phase rather than treating environmental impact as an afterthought. The company developed a Sustainable Portfolio Management tool that evaluates products on both technical performance and environmental footprint.

AI accelerating materials discovery

The relationship between AI and materials science runs in both directions. Syensqo uses AI agents to digitally synthesize millions of potential molecular combinations, predict their performance and sustainability characteristics, and identify promising candidates for laboratory testing. This approach allows researchers to explore possibilities "broader, deeper, and faster" while focusing human expertise on complex engineering problems.

Finelli envisions a reinforcing cycle: AI helps develop materials that improve AI infrastructure, which enables better AI systems to accelerate materials discovery. "You end up in this accelerated materials, innovative cycle of materials innovation," he said.

Why it matters

As AI systems grow more powerful and energy-intensive, the materials that enable semiconductor manufacturing and data center operations are becoming strategic constraints. Companies that can solve materials challenges around thermal management, electrical efficiency, and sustainability will shape the boundaries of what AI systems can achieve. The convergence of AI-driven materials discovery and infrastructure demands could determine which organizations lead the next phase of computing innovation.

These details were first reported by MIT Technology Review in a Business Lab podcast episode produced in partnership with Syensqo.