List-price math only. Real Snowflake bills can be cut 40 to 70 percent by tuning auto-suspend per warehouse, right-sizing for actual workload, and moving non-regulated workloads off Business Critical. The deep audit models your real QUERY_HISTORY and WAREHOUSE_METERING_HISTORY to rank reduction wins by dollar impact.
| Line item | Math | Monthly |
|---|---|---|
| — | ||
| Warehouse size | Credits / hr | $/hr (Standard) | $/hr (Enterprise) | $/hr (Business Critical) |
|---|---|---|---|---|
| X-Small | 1 | $2.00 | $3.00 | $4.00 |
| Small | 2 | $4.00 | $6.00 | $8.00 |
| Medium | 4 | $8.00 | $12.00 | $16.00 |
| Large | 8 | $16.00 | $24.00 | $32.00 |
| X-Large | 16 | $32.00 | $48.00 | $64.00 |
| 2X-Large | 32 | $64.00 | $96.00 | $128.00 |
| 3X-Large | 64 | $128.00 | $192.00 | $256.00 |
| 4X-Large | 128 | $256.00 | $384.00 | $512.00 |
One-page checklist of the biggest Snowflake levers — auto-suspend tuning by workload type, warehouse right-sizing using QUERY_HISTORY, the multi-cluster ceiling trap, when to switch off Business Critical, query result caching, materialized view break-even, clustering keys vs auto-clustering cost, and the Snowpipe streaming pricing mistake that doubles ingest bills. PDF sent to your inbox.
credits_per_query = warehouse_size × (runtime + auto_suspend) / 3600
Example: X-Small warehouse (1 credit/hr) × 30s runtime + 60s auto-suspend = 90 effective seconds per query. 1 × 90 / 3600 = 0.025 credits per query. 1,000 queries/day × 30 days = 750 credits/month. At Standard edition $2/credit = $1,500/month. The same workload without auto-suspend overhead (1 × 30 / 3600 = 0.00833 credits/query × 30,000/month × $2) drops to $500/month — a 67 percent reduction.
Scale to a Medium warehouse with the same defaults and the math becomes 4 × 90 / 3600 = 0.1 credits/query × 30,000 queries/month × $2 = $6,000/month. Switch to Business Critical and that same workload jumps to $12,000/month with zero performance difference. The two biggest reductions on most teams: drop auto-suspend from 60 to 10 seconds (saves 40-55 percent of warehouse time on bursty workloads), and downsize one tier where queries don't actually saturate the larger warehouse (saves 50 percent linearly).
Snowflake bills for warehouse time, not query work. Once a warehouse is running, credits accrue every second whether queries are executing or the cluster is idle waiting for the next one. The credit rate doubles with every warehouse size, and at Business Critical's $4 per credit, an idle X-Large warehouse costs $64 per hour. The other quiet driver is auto-suspend — the default of 60 seconds of idle before suspend means every query effectively pays for at least one extra minute. Teams that don't tune size, suspend interval, and edition typically overpay by 2-3x what their actual query work demands.
Auto-suspend tells the warehouse how long to wait after the last query before shutting itself down. Lower values save money because billing stops the moment the warehouse suspends. The default 60 seconds is conservative — it covers BI dashboards where users click around and queries fire 10-30 seconds apart. For batch ETL or ad-hoc analyst queries where the next query is minutes away, drop auto-suspend to 30 or 10 seconds. A workload with 1000 queries/day at 30s runtime pays for 1000 × 90 = 25 hours of warehouse time, but the actual work is only 8.3 hours. Drop suspend to 10 and that becomes 11.1 hours — a 56 percent reduction.
Standard is $2/credit, Enterprise $3, Business Critical $4. The edition you need depends on features, not query speed — queries run identically across editions on identical warehouse sizes. Enterprise unlocks materialized views, 90-day time travel, column-level security, multi-cluster warehouses, and dynamic data masking. Business Critical adds HIPAA / PCI compliance, customer-managed keys, and private connectivity. Healthcare and fintech teams need Business Critical's compliance posture. General SaaS teams needing multi-cluster concurrency want Enterprise. Everyone else should start on Standard. The most common waste pattern: teams running on Business Critical out of habit while only using Standard features.
Scale UP (bigger warehouse size) when individual queries are slow. Doubling warehouse size typically halves query runtime on data-intensive work — same credit cost, better wall-clock. Scale OUT (multi-cluster auto-scaling, Enterprise+) when concurrency is the problem. Multi-cluster spins up additional clusters when queries queue, each billing independently. Set max_clusters=3 and you can pay up to 3x your single-cluster rate during spikes. The trap: leaving max_clusters at 10 with a flaky workload can quietly multiply your bill by ten overnight. Scale UP for slow queries; scale OUT only if concurrency truly demands it AND set a tight max_clusters ceiling.
No. Warehouse sizes, runtimes, query counts, and rate math all run locally in your browser. The page fires an anonymous pageview beacon and CTA-click events so we can measure whether the calculator is useful — no inputs, no email (unless you submit one to the cheat-sheet form), no IP stored raw.