An e-commerce company runs nightly 5-hour inventory reconciliation on x86 instances and processes order events through orchestrated workflows with millions of sub-second checks plus occasional multi-hour exception handling flows. Compute dominates the bill and transitions are charged per state change. Which pricing-model architecture minimizes cost for this workload mix?

The correct architecture minimizes both cost drivers by combining Graviton with Compute Savings Plans for reconciliation and splitting workflows by billing model: Graviton provides up to 40% price-performance improvement for compatible workloads and Savings Plans discount predictable recurring compute versus On-Demand, critical when compute dominates the nightly 5-hour job, while Standard per-transition pricing fits infrequent long-running exception flows with full history and exactly-once semantics and Express duration-plus-invocations billing is dramatically cheaper for millions of sub-second checks. The Standard-only workflow design fails on workflow economics by forcing high-volume short checks onto per-transition pricing, inflating cost by orders of magnitude where Express would aggregate work. The Graviton On-Demand split-workflow design fails on commitment economics by omitting Savings Plans for a fixed nightly job, leaving a guaranteed discount uncaptured for no architectural benefit. The Express-only workflow design fails on service limits because Express synchronous and asynchronous executions cap at 5 minutes and cannot support occasional multi-hour flows requiring Standard up to one year. Exam caveat: do not default to consistency or a single workflow type when the stem explicitly contrasts millions of short executions versus few long executions and calls out per-transition charges, as that contrast signals a hybrid Express plus Standard split. Practical check: verify Graviton compatibility and enable Compute Optimizer plus Cost Explorer, then tag Standard versus Express executions and compare transition versus duration charges before committing Savings Plans coverage to the steady-state baseline.