Snowflake is consumption-based: you buy credits for compute (per-second, 60-second minimum) and pay separately for storage. The bill runs away through idle warehouses, oversized compute, and staying on-demand instead of a discounted capacity commitment.
What you should really pay
| Item | Benchmark |
|---|---|
| Standard / Enterprise credit | ~$2.00 / ~$3.00–4.65 |
| Storage — on-demand vs capacity | ~$40 vs ~$23 / TB / mo |
| Capacity commitment minimum | ~$25,000 (drawn down) |
The cost traps
- Idle warehouses / bad auto-suspend — the #1 leak; a left-running large warehouse burns cash fast.
- Oversized warehouses — each size up doubles credit burn.
- Serverless features (Search Optimization, Auto-Clustering) accruing silently.
- Commitment shortfall — unused committed credits are largely use-it-or-lose-it.
The levers that work
- Right-size first: fix auto-suspend and warehouse sizing before you commit, so you don't over-commit.
- Capacity commitment discounts: ~10–15% ($0.5–1M), ~20–25% ($1–5M), ~30–35% ($5M+); multi-year adds ~5–10%.
- Get written rollover terms for unused credits.
- Route via your cloud Marketplace to retire committed cloud spend.
- Bring a live POC on Databricks or BigQuery — the strongest single lever, paired with the January fiscal-year-end.
Bottom line
Optimize consumption first, then commit at the right tier with rollover protection. Right-sizing before you sign is what keeps you from over-committing.
Get the full Snowflake playbook
Real pricing benchmarks, every cost trap with dollar examples, the discount levers and the ranges buyers actually hit, fiscal-year timing, and copy-paste negotiation emails.
Browse the benchmarksGet the free checklistFigures are estimated market ranges compiled from public and third-party sources to help you anchor a negotiation; they are not legal, financial, or procurement advice and are not affiliated with or endorsed by any vendor named. Always confirm against your own quote.