Updated 19 Oct 2026 • 5 mins read
Snowflake Cost Calculator: Estimate Credits, Storage, and Warehouse Spend
This free Snowflake cost calculator estimates warehouse credits by size and running time, including the idle hours before auto-suspend, plus storage, serverless features, cloud services, and Cortex AI Credits, at 2026 rates by edition. The page explains why credits are hard to forecast, works an example, and shows how to replace the estimate with real spend.
Snowflake is one of the few bills that is harder to forecast than it is to read. The rate card is short, a credit price by edition and a doubling table of warehouse sizes, and yet the question what will this cost next month defeats most data teams, because credits are consumed by time rather than by work. A warehouse bills for every second it runs, including the minutes it idles waiting for auto-suspend; a feature someone switched on in a sandbox bills around the clock; and since April, AI calls bill in a second currency. The estimate built from job schedules is right about the queries and wrong about everything between them.
The calculator below estimates a Snowflake account the way Snowflake bills it: each warehouse by size, active hours, and idle hours, at your edition's credit price, plus storage, serverless features, cloud services above the free allowance, and Cortex AI Credits on their own line, with the idle share reported so the usual blind spot is visible. The page around it explains the meters, works an example, and shows how to replace the estimate with the spend actually billed. The levers that reduce the result are in our Snowflake cost optimization guide; the AI meter is in our guide to Snowflake Cortex pricing. Rates are from Snowflake's published pricing as of October 2026.
The Four Meters on a Snowflake Bill
| Meter | How it bills | 2026 rate (AWS US East) | What estimates usually miss |
|---|---|---|---|
| Warehouse compute | Credits per second a warehouse runs, 60-second minimum per resume, by size | 1 credit/hour at X-Small doubling to 128 at 4X-Large; $2, $3, $4, or about $6 per credit by edition | Idle time before auto-suspend, often 20 to 40 percent of credits |
| Storage | Compressed TB per month | $23 on capacity, $40 on demand | Time Travel and Fail-safe copies on heavily updated tables |
| Serverless and cloud services | Credits for features that run without a warehouse; cloud services free to 10 percent of daily compute | Search optimization and MV refresh about 10 credits/hour each; clustering; Snowpipe per file | Features enabled once and never audited |
| Cortex AI | AI Credits, a separate currency since April 1, 2026 | $2.00 global routing, $2.20 home region | Warehouse compute the same query also wakes |
The first row is the one that decides most estimates. Snowflake bills a warehouse from the moment it resumes to the moment auto-suspend fires, and the default auto-suspend of five or ten minutes means a warehouse hit by a query every eight minutes never suspends at all. The calculator asks for idle hours separately so that the number is estimated rather than assumed to be zero.
A Worked Example
The calculator's defaults describe a typical Enterprise-edition account with four warehouses: a Large for ETL running 180 hours a month with 20 idle, a Medium for BI running 400 active and 120 idle, a Medium for data science at 150 and 60, and a Small for ad hoc work at 100 and 80, with 20 TB of storage on capacity pricing.
Warehouse credits come to 4,880 a month, about $14,640 at $3 a credit, plus $460 of storage: roughly $15,100. The line most teams would not have estimated is the idle share: 1,040 of those credits, about $3,120 a month, are warehouses running with nothing to do, 21 percent of compute. Set auto-suspend to 60 seconds on all four and most of it goes; step the BI warehouse from Medium to Small for routine dashboards and the BI line halves. The calculator cannot tell you which warehouses are oversized, only what they cost at the size they are, which is the point where an estimate stops and measurement starts.
Why Credits Resist Forecasting
- Time, not work. Two identical query loads cost different amounts depending on how long their warehouses idle, which depends on auto-suspend settings and query timing nobody budgets.
- Size is a decision, not a fact. A warehouse left at Large because the month-end report needed it runs every routine query at Large rates. The estimate prices the size; the waste is the mismatch.
- Meters that start silently. Search optimization, materialized views, and clustering bill continuously from the moment someone enables them in a sandbox and forgets.
- Concurrency and multi-cluster. Multi-cluster warehouses add clusters under load and bill each one; peak concurrency is rarely in the estimate.
- The second currency. Cortex AI Credits bill the inference and the warehouse the query woke, on two meters, which a sizing estimate does not model.
Each of these is a reason the actual bill drifts from the estimate month over month, and the reason our Snowflake cost optimization guide puts auto-suspend and right-sizing ahead of every other lever.
From Estimate to Real Spend
The button under the result exists because the hard part of Snowflake cost is not the arithmetic but the inputs: real running hours, real idle time, which features are on, and which teams and customers are consuming each shared warehouse. Opslyft ingests Snowflake usage natively and shows warehouse-level spend in dollars at your credit price in Explorer, beside your AWS, Azure, and GCP bill, broken down by team, project, or customer and tracked over time, with query-level attribution for the shared warehouses that an estimate can only treat as one line. Dimension Studio allocates warehouse credits to the teams and customers whose queries ran, Atlas lists every warehouse with an owner and a cost, and Monitor catches the warehouse resized up and left there, the serverless feature enabled on a large table, and the Cortex job that looped, with the cause, before the month closes. Forecasts then run on allocated actuals rather than on a sizing guess, which is the difference between an estimate and a number finance can plan on.
How to Pay Less for Snowflake
- Set auto-suspend to 60 seconds on every warehouse; the calculator's idle line is the saving.
- Right-size one step at a time, keeping a larger warehouse only for the queries that need it.
- Separate ETL, BI, data science, and ad hoc onto their own warehouses so each is sized and suspended for its own pattern.
- Audit serverless features quarterly and disable what is not earning its credits.
- Cap non-production with resource monitors.
- Negotiate a capacity contract once usage is stable; 20 to 30 percent off the credit rate.
- Attribute warehouse credits to the teams and customers that consume them, because a warehouse nobody owns is a warehouse nobody suspends.
For how Snowflake compares with the cloud-native warehouses, see our comparison of Snowflake, AWS, and Azure; for the anomaly patterns specific to Cortex, our guide to Snowflake AI agent cost anomalies.
FAQs
Add four meters. Warehouse compute: each warehouse's credits per hour (1 at X-Small, doubling per size to 128 at 4X-Large) times the hours it runs, including idle time before auto-suspend, times your credit price ($2, $3, or $4 by edition in AWS US East, about $6 for VPS). Storage: compressed terabytes times $23 (capacity) or $40 (on demand). Serverless features and cloud services above the free 10 percent, in credits. And Cortex AI Credits at $2.00 to $2.20 on their own meter. The calculator above does this per warehouse.
About $2.00 on Standard, $3.00 on Enterprise, $4.00 on Business Critical, and about $6.00 on Virtual Private Snowflake, on demand in AWS US East. Other regions run 30 to 60 percent higher, and capacity contracts bring the rate down 20 to 30 percent. Since April 1, 2026, Cortex AI features bill in a separate AI Credit priced at $2.00 with global routing or $2.20 pinned to your home region.
One credit per hour at X-Small, doubling with each size: Small 2, Medium 4, Large 8, X-Large 16, 2X-Large 32, 3X-Large 64, 4X-Large 128. Billing is per second with a 60-second minimum each time a warehouse resumes, and the warehouse bills for every second it is running, including the idle time before auto-suspend fires. A Large warehouse idle for an hour costs 8 credits, about $24 on Enterprise.
Because credits are consumed by time, not by query, and time depends on behavior nobody budgets: how long warehouses idle before suspending, which size each team leaves its warehouse at, which serverless features were switched on, and how many Cortex calls a new feature makes. A forecast built from warehouse sizes alone misses idle time, which is often 20 to 40 percent of credits, and the serverless and AI meters, which bill around the clock once enabled.
Yes, as its own line. Cortex functions, Search, Analyst, and Agents bill in AI Credits, a separate currency from the platform credits warehouses consume, at $2.00 or $2.20 per AI Credit since April 2026. Enter your expected monthly AI Credits; the warehouse compute a Cortex query also consumes is captured in the warehouse rows.
As accurate as the running-hours and idle-hours figures you enter, which is the hard part. Teams that measure actual warehouse uptime from Snowflake's account usage views, including idle time, typically estimate within 10 percent; teams that estimate from job schedules alone usually land 20 to 40 percent low. The calculator reports the idle share so the gap is visible.