Updated 18 Oct 2026 • 7 mins read
Snowflake Cost per Query and per Credit: See Warehouse Spend Next to Your Cloud Bill in Opslyft
Snowflake bills in credits by warehouse time, invoices separately from the clouds, and sits outside most cost tools, so the second-largest bill in many data-heavy companies is reviewed once a year. This walkthrough shows how Opslyft ingests it natively, converts credits to dollars, attributes cost per query and per customer, and catches runaway queries.
Ask a data-heavy company what its second-largest infrastructure bill is and the answer is usually Snowflake, and the follow-up, what is the Snowflake cost per query, is usually met with silence, because Snowflake does not bill per query. It bills in credits, by warehouse size and running time, invoices separately from AWS, Azure, and GCP in a currency the cloud cost tool does not read, and reports usage in its own account views. So the warehouse that serves a dozen teams and every customer dashboard is budgeted by the data platform lead, reviewed at renewal, and never attributed to the teams, queries, or customers that actually consumed it, while the compute and storage that feed it are allocated to the last dollar on the cloud bill next door.
This page is a walkthrough of how Opslyft closes that gap: ingesting Snowflake natively, converting credits to dollars at your edition's rate, showing warehouse spend in the same view as AWS, Azure, and GCP, attributing it per query, per team, and per customer with the same rules as cloud spend, and catching the runaway query in hours rather than on the invoice. The optimization levers themselves are in our Snowflake cost optimization guide; the separate AI meter is in our guide to Snowflake Cortex pricing; this page is about seeing the bill.
The short answer: Snowflake bills warehouse compute in credits by size and running time, so cost per query is derived rather than billed: the credits a warehouse consumed during a query, attributed from Snowflake's query history and warehouse metering, multiplied by your credit price, about $2, $3, or $4 on Standard, Enterprise, and Business Critical in AWS US East. Because Snowflake invoices separately from the clouds in its own currency, most cloud cost tools never see it, and the second-largest bill in many data-heavy companies goes unattributed. Opslyft ingests Snowflake natively: Explorer shows warehouse-level spend in dollars beside AWS, Azure, GCP, OCI, and Kubernetes, by team, project, or customer; Dimension Studio attributes shared warehouses per query, role, and customer by the same rules as cloud spend; Monitor catches a warehouse or query anomaly with its cause; and the Glue jobs and S3 buckets that feed the warehouse appear next to it, so a data product's full cost is one number.
Why Snowflake Lives Outside Your Cost Tool
Three structural reasons, none of them Snowflake's fault.
A different invoice in a different currency. Snowflake bills credits, priced by edition, on its own invoice; cloud cost tools read AWS, Azure, and GCP billing exports. Nothing on the cloud bill says what the warehouse cost, and nothing on the Snowflake bill says which team used it.
Compute billed by time, consumed by query. A warehouse bills credits for every second it runs, 60 seconds minimum per resume, serving many queries from many teams concurrently. The invoice knows the warehouse's credits; the account usage views know the queries; joining them with concurrency and idle time accounted for is real work.
A second meter since April. Cortex AI features bill in AI Credits at $2.00 to $2.20, separate from platform credits, which means a data product's Snowflake cost is now two currencies on one invoice.
The consequence is familiar from every other service that bills by consumption: the warehouse that costs the most is the one nobody owns, and the query that costs the most is the one nobody can see.
| Edition | Credit price (AWS US East, on demand) | X-Small warehouse per hour | Large warehouse per hour | Large warehouse idle for one hour |
|---|---|---|---|---|
| Standard | About $2.00 | $2 | $16 | $16 for nothing |
| Enterprise | About $3.00 | $3 | $24 | $24 for nothing |
| Business Critical | About $4.00 | $4 | $32 | $32 for nothing |
How Opslyft Solves It
Step 1: Snowflake is a first-class platform, not an import
Opslyft ingests Snowflake usage natively, alongside AWS, Azure, GCP, OCI, and Kubernetes, and Explorer shows warehouse-level spend in dollars at your edition's credit price, in the same view as the cloud bill, broken down by team, project, service, or customer and tracked over time. Platform credits and Cortex AI Credits appear as what they are, two meters on one data product. The Glue or Data Factory jobs that load the warehouse and the S3 or Blob storage behind external tables sit beside it, so the data platform's cost is one picture rather than three invoices reconciled in a spreadsheet.
Step 2: Cost per credit becomes cost per query
Snowflake's account usage views record every query with its warehouse, user, role, duration, and bytes scanned, and the warehouse metering history records the credits each warehouse consumed per hour. Opslyft joins them, attributing a warehouse's credits to the queries that ran on it with concurrency and resume minimums accounted for, and converts the result to dollars. The output is cost per query, per user, per role, and per warehouse, which is the view that turns the silence after what does this query cost into a number, and the top twenty queries by credits into a list someone can fix.
Step 3: Per query becomes per team and per customer
Dimension Studio maps Snowflake spend to the same teams, products, and customers as cloud spend, by warehouse where teams have their own and by query attribution where they share, splitting a shared BI warehouse's credits across the roles and customers whose queries ran on it by rule, fixed, proportional, or usage-based. For a multi-tenant data product this is the step that makes cost per customer include the warehouse credits that customer's dashboards drove, the number that connects a Snowflake bill to a margin, and it is what lets finance and the data team argue about priorities rather than arithmetic.
Step 4: Every warehouse has an owner, and a history
Atlas inventories every warehouse with an owner and a cost, so the X-Large that a team created for a migration eighteen months ago has a name next to it and a change history that shows who resized it last. Its recommendations weigh what a warehouse is for, so an ETL warehouse and an ad hoc one at the same credit burn get different advice: a schedule and a suspend setting for one, a resource monitor for the other.
Step 5: The runaway query is caught in hours
Monitor detects Snowflake anomalies at warehouse and query level with a root-cause explanation, a warehouse resized up and left there, search optimization enabled on a large table, a Cortex job looping, an unbounded embedding query against a 41-million-row table, and routes them to the data team's Slack channel; budgets per warehouse, team, or account hold the result. Opslyft documented exactly that last case: 2,340 credits in six hours from one unbounded Cortex query, 19 percent of a customer's monthly Snowflake budget, the pattern our guide to Snowflake Cortex spend anomalies describes. Tickets turns the fix into an assigned item in Jira, ServiceNow, or Asana tracked to a measured saving, and Iris answers which teams' warehouse spend grew fastest this month and why, in plain English.
A Worked Example
An Enterprise-edition account spends about $16,000 a month on Snowflake across four warehouses, invoiced separately from an AWS bill of $140,000 that the FinOps team allocates to the last dollar. The data platform lead owns the Snowflake number; nobody owns the warehouses inside it, and the Large BI warehouse shared by six product teams and every customer dashboard is one line.
On Opslyft, the $16,000 appears in Explorer beside the $140,000, and the shared BI warehouse resolves into its queries: the top twenty by credits account for 44 percent of it, and three of them are customer-facing dashboards rescanning an unclustered 2 TB table every hour. Per-customer attribution shows two enterprise customers driving 31 percent of all warehouse credits on a flat-rate contract. Monitor flags a fourth warehouse, the migration X-Large, resuming nightly for a job whose consumer was retired. The data team clusters the table and fixes the three dashboards, saving about $2,400 a month; the X-Large is dropped, saving $900; and the two enterprise customers' consumption goes into the next renewal conversation with a number attached. Snowflake is now allocated like AWS, by the same team, against the same budgets, and the next anomaly reaches the owner in hours.
Proof: what customers say, and the Innovaccer result
“OpsLyft has given us clarity in our cloud spend across AWS, Azure, and Snowflake,” one customer wrote in a G2 review. Innovaccer, running one cloud account per customer across AWS, Azure, GCP, and Snowflake, brought warehouse spend into the same accountability model as everything else, tied each customer's cloud budget to that customer's MRR, and improved its MRR-to-cloud-cost ratio by 35 percent while cutting cloud cost by 30 percent. Read the full story: How Innovaccer improved its MRR-to-cloud-cost ratio with Opslyft
Use Opslyft If
Snowflake is one of your largest bills and it is reviewed once a year by one person.
Shared warehouses serve many teams or customers and nobody can say who consumed what.
Finance needs cost per customer and the warehouse credits are missing from it.
The Snowflake invoice and the cloud invoice are reconciled by hand, or not at all.
Cortex AI Credits have appeared on the bill and nobody has attributed them.
Runaway queries are discovered on the invoice rather than in the hour they run.
You want Snowflake governed by the same budgets, alerts, and tickets as AWS, Azure, and GCP.
If Snowflake is a small line with one warehouse and one owner, Snowflake's own account usage views and a resource monitor will do. The platform earns its place when the warehouse is shared, the bill is large, and it lives outside everything else you manage.
The Bottom Line
Snowflake cost per query is a number Snowflake does not bill and every data-heavy company needs, because the warehouse that serves everyone is the one nobody owns and the query that costs the most is the one nobody sees. Opslyft ingests Snowflake natively, converts credits to dollars, shows warehouse spend beside the cloud bill, attributes it per query, team, and customer by the same rules as everything else, and catches the runaway query in hours. The second-largest bill stops being a line item reviewed at renewal and becomes part of the same practice as the first. That is what Opslyft's cost visibility is built for across every platform you pay for; for how Snowflake compares with the cloud-native warehouses, see our comparison of Snowflake, AWS, and Azure.
FAQs
Snowflake bills warehouse compute in credits by warehouse size and running time, not per query, so per-query cost is derived: the credits a warehouse consumed during a query's execution, attributed to that query from the account usage views (query history and warehouse metering), multiplied by your credit price, about $2 on Standard, $3 on Enterprise, and $4 on Business Critical. Because warehouses serve many queries at once and bill a 60-second minimum per resume, attribution has to account for concurrency and idle time, which is what makes it hard to do in a spreadsheet.
The dollar price of one Snowflake credit, set by edition, cloud, and region, and by whether you pay on demand or on a capacity contract. In AWS US East on demand it is about $2.00 on Standard, $3.00 on Enterprise, $4.00 on Business Critical, and about $6.00 on Virtual Private Snowflake; other regions run 30 to 60 percent higher and capacity contracts 20 to 30 percent lower. Since April 2026, Cortex AI features bill in separate AI Credits at $2.00 to $2.20.
Because Snowflake invoices separately from AWS, Azure, and GCP, in its own currency, from its own usage views, and most cloud cost tools read only cloud billing exports. The data warehouse ends up budgeted by the data team, reviewed annually, and never reconciled with the compute, storage, and pipelines on the cloud bill that feed it, even though the same teams and customers drive both.
Opslyft ingests Snowflake usage natively and shows warehouse-level spend in Explorer in the same view as AWS, Azure, GCP, OCI, and Kubernetes cost, broken down by team, project, or customer and tracked over time. Credits are converted to dollars at your edition's rate, Cortex AI Credits are shown beside platform credits, and the Glue or Data Factory jobs and S3 or Blob storage that feed the warehouse appear next to it, so a data product's full cost is one number.
Yes. Query-level attribution uses Snowflake's own query history and warehouse metering to assign credits to the queries, roles, and users that consumed them, and Dimension Studio maps those to teams, products, and customers by the same rules as cloud spend, splitting shared warehouses by actual query share. That is how cost per customer includes the warehouse credits that customer's usage drove, which is the number that connects a Snowflake bill to a margin.
Monitor detects Snowflake anomalies at warehouse and query level with a root-cause explanation, a warehouse resized up and not down, a serverless feature enabled on a large table, a Cortex job looping, and routes them to the data team's Slack channel before the month closes. Opslyft documented one unbounded Cortex embedding query that consumed 2,340 credits in six hours, 19 percent of a customer's monthly Snowflake budget; that is the class of event per-query attribution is built to surface in hours rather than on the invoice.
Serverless features that bill continuously once enabled, search optimization and materialized-view refresh at roughly 10 credits an hour each, automatic clustering on high-churn tables, and Snowpipe per file; the 60-second minimum on every warehouse resume; cloud services above the free 10 percent of daily compute; Time Travel and Fail-safe storage on heavily updated tables; Cortex AI Credits; and cross-region egress billed by the cloud provider. Each lands on the Snowflake invoice as a line most teams never break down.