I built this tool because comparing cloud GPU pricing to buying hardware has become increasingly messy for AI workloads. I wanted a way to input real workload parameters (tokens/sec, request patterns, model size) and see where the break-even point actually is.

How it works: it models throughput and utilization, then compares current cloud pricing (AWS/GCP) against ownership costs for on-prem GPU clusters over time.

Free/Paid: the online diagnostic is free and gives the core recommendation. I charge $99 for a detailed PDF with 36-month cash-flow projections and charts, mainly for teams that need to justify decisions to a board or CFO.

I’d really appreciate feedback on the assumptions and modeling approach.