ENEOS accelerates AI-driven materials discovery of immersion cooling fluids and novel catalysts Using NVIDIA ALCHEMI
ENEOS Holdings, Inc. announced that it has implemented AI-driven chemistry and materials discovery platform, NVIDIA ALCHEMI, to accelerate identification and optimization of next‑generation immersion cooling fluids1 and oxygen evolution reaction (OER) catalysts.
ENEOS carried out large-scale virtual screenings for two use cases leveraging NVIDIA ALCHEMI, a platform that accelerates chemistry-specific computational workloads. This approach allows for more efficient simulations and more effective chemical representations of formulations using AI, compared to traditional methods. For immersion cooling fluids aimed at data center thermal management, the workflow screened on the order of ten million molecular candidates against dielectric and thermal properties and identified approximately 1,000 molecules that met or exceeded the target performance envelope for further experimental evaluation. For oxygen evolution reaction (OER) catalyst discovery seeking low-cost alternatives to iridium oxide, the computational funnel evaluated candidate spaces approaching 100 million material hypotheses and produced approximately 1,000 promising catalyst compositions for prioritized synthesis and testing.
These results reflect an overall reduction in the discovery timeline, from a process that historically required years to be completed in months with accelerated simulation, machine learning interatomic potentials (MLIPs), and automated workflow orchestration. Specifically, ENEOS HD led the design of the screening objectives and the selection criteria that translated application-level requirements - such as thermal, dielectric properties and catalytic activity - into computationally tractable metrics. This domain-driven orchestration allowed the team to focus computational and validation resources on candidates most likely to deliver real-world performance. By leveraging NVIDIA ALCHEMI, ENEOS accelerated calculations of Matlantis’s core technology, PreFerred Potential (PFP), by an order of magnitude — that is, several tens of times faster.
1 Announced on March 31, 2025
Comments