This article explains best-practices for building enterprise-grade HPC infrastructure on AWS, emphasizing the importance of foundational "plumbing" and shared responsibility models.

• HPC success depends on solid foundational infrastructure, not just compute resources
• Shared responsibility model: collaborate with central IT rather than building everything independently
• AWS Landing Zones provide pre-defined, well-architected multi-account enterprise infrastructure
• Landing Zone Accelerator (LZA) enables repeatable configuration through infrastructure-as-code YAML
• Landing Zones include security, shared services, and workload account organizational units
• HPC teams build within workload accounts while leveraging shared services and guardrails
• Cloud elasticity allows compute to scale up/down based on demand, reducing costs
• Service Catalog enables controlled self-service infrastructure provisioning for end-users
• Consolidated billing across accounts enables better Savings Plans and volume discounts
• Separate workload accounts simplify cost tracking and charge-back processes

Enterprise HPC on AWS succeeds through collaboration with central IT services, leveraging Landing Zones as foundational infrastructure, and embracing cloud elasticity for cost optimization and user flexibility.