How Aionics accelerates chemical formulation and discovery with AWS Parallel Computing Service

This article describes how Aionics uses AWS Parallel Computing Service (PCS) to accelerate chemical formulation and battery electrolyte discovery through high-performance computing.

• Aionics migrated from AWS ParallelCluster to AWS PCS in August 2025 for managed cluster operations
• Deployed two separate PCS clusters: CPU cluster (hpc6a/hpc7a instances) and GPU cluster (g6e instances)
• hpc7a.48xlarge delivers 40% faster performance than hpc6a.48xlarge at 40% higher cost
• hpc7a provides 8 GiB memory per core versus 4 GiB, enabling larger systems per node
• Larger GPU instances (g6e.48xlarge) cost 50% less per system than smaller instances for molecular dynamics
• Uses Easybuild framework for managing complex Fortran/C++ software dependencies via Environment Modules
• Multiple queues enable optimization for different objectives: cost efficiency versus faster completion
• EC2 Capacity Blocks for ML integration enables GPU reservation up to 8 weeks in advance

AWS PCS provides flexible, managed HPC infrastructure enabling Aionics to focus on battery research rather than infrastructure maintenance while optimizing for cost or performance based on workload requirements.