Methodology

Each estimate starts with a published vendor reference architecture. A size tier maps to a concrete set of components, those components are matched to cloud instance shapes, and cached pricing snapshots provide the monthly cost.

Precomputed AI design pattern

CloudEstimate is a worked example of the Precomputed AI design pattern for artifact-first LLM systems. Reference architectures, shape mappings, pricing snapshots and generated explanation caches are versioned artifacts produced ahead of time and served without live model inference. Snapshot dates are shown on every page because dated numbers are more defensible than implied recency.

Methodology citation: Raquedan, R. (2026). Precomputed AI: Reason Ahead of Time, Serve Instantly. https://precomputedai.com.

What is modelled

• VM instance sizing by component role
• Persistent storage sized from the component footprint
• Region and commitment-term differences for commercial pricing
• High-availability overhead when the source architecture requires extra nodes

What is not modelled

• Vendor licensing and support contracts
• Network egress, cross-region traffic, and DR topologies
• Compliance controls, backups, and monitoring platforms
• Managed-service alternatives and hybrid-use benefits

Refresh cadence

Pricing snapshots are intended to refresh nightly. Reference architectures refresh when vendors publish materially changed guidance. Dates are always shown on-page because dated numbers are more defensible than implied recency.

Reporting errors

If a route looks wrong, open an issue with the estimate URL, the cloud, the region, and the pricing source that disagrees. Contributions for new ISVs should follow the YAML schema and validation flow in the repository.