Snowflake Openflow pricing in instance hours, and the floor you pay before a row moves
7 min read Cost The Adapters team
Field mapping auto-plugged · tap a port to rewire
Plug a source port into
Transform on this cable
JSON in
JSON out
5 sample records ready
Snowflake does not publish a price for Openflow connectors. In a Snowflake deployment you pay compute pool credits per instance hour, with a 5 minute minimum and a control pool that runs even when no connector does. For one small runtime that is 730 instance hours a month before a single row moves, and 1,460 if the runtime never stops.
Key takeaways
- The unit is an instance hour. Not a row, a sync or a connector.
- The floor is the control pool. It runs with no connector active.
- BYOC adds a second bill. Your cloud provider charges for the infrastructure.
How much does Snowflake Openflow cost?
The honest answer is that it depends on a rate Snowflake keeps in its service consumption table, which is tied to your contract. What the Openflow cost page for Snowflake deployments does state is the shape of the bill. Compute pools are charged in credits per instance hour, billed per second with a 5 minute minimum. A Small runtime maps to a CPU_X64_S instance, a Medium to CPU_X64_SL and a Large to CPU_X64_L. And a control pool of one CPU_X64_S instance runs even with no runtimes, which the page shows in its own example. We read it on 20 November 2026.
What a month of instance hours looks like
The same page works one small runtime for one hour as 2 instance hours, one for the control pool and one for the compute pool. Scale that to a 730 hour month and the floor is plain. Whatever the credit rate, the shape below is what you are multiplying it by.
| Setup | Instance hours a month | How we got it |
|---|---|---|
| Control pool only | 730 | 1 instance for 730 hours |
| Small runtime, 1 hour a day | 760 | 730 + 30 |
| Small runtime, 4 hours a day | 852 | 730 + 122 |
| Small runtime, 8 hours a day | 973 | 730 + 243 |
| Small runtime, always on | 1,460 | 730 + 730 |
Two lessons come out of it. First, an hour of runtime a day adds only 30 hours to the 730 hour floor, so shortening the schedule saves far less than people expect. Second, an always-on runtime doubles the month. Whether a given connector can stop between runs is worth asking your Snowflake account team before you plan on the cheap row.
Snowflake deployment, BYOC or a flat plan
The BYOC cost page describes the other deployment. It bills vCPU usage per second with a 60 second minimum and leaves the underlying cloud infrastructure to your cloud provider, so you add two invoices. The table sets the three side by side.
| Route | Billing unit | Minimum | Also pay for | Watch for |
|---|---|---|---|---|
| Snowflake deployment | Compute pool credits per instance hour, by instance family | 5 minutes | None, it is all Snowflake credits | Control pool runs with no runtimes, no per-runtime cost split |
| BYOC deployment | vCPU usage, billed per second | 60 seconds | Your own cloud bill for the infrastructure | Two invoices to add together |
| Flat priced connector | One monthly plan, no instance hours | None | The Snowflake warehouse that receives the rows | Not a Snowflake-managed connector |
How do I see what Openflow costs in my account?
Do this before you sign off on a pilot. On a Snowflake deployment the view OPENFLOW_USAGE_HISTORY returns no rows. Cost shows up in METERING_HISTORY with the service type OPENFLOW_COMPUTE_SNOWFLAKE, and the column that normally names the service shows the compute pool that incurred it. Because runtimes of one size in a deployment share a pool, you cannot split the cost per connector, so a pilot on a dedicated deployment gives the cleanest number.
Is Openflow cheaper than a third-party ETL tool?
The first-party route wins when you already run many Openflow connectors on shared runtimes, because the floor is spread across all of them, or when you want everything inside one Snowflake contract. A flat plan wins for one or two sources with small daily change volume, where the floor is most of the bill. The Shopify case is worked through on Shopify to Snowflake integration tools, and the per-schedule warehouse cost on the receiving side is on QuickBooks to Snowflake integration tools. For CRM sources, compare it with the best HubSpot to Snowflake connector and Salesforce to Snowflake connector pricing. Stores that publish many near-duplicate collection pages sometimes pair a data pipeline with a service for removing stale pages from search results so reports and indexed URLs agree.
Do I pay for Openflow when no connector is running?
In a Snowflake deployment, yes. The cost page shows a control pool of one small instance that runs even with no runtimes, which is 730 instance hours in a 730 hour month. Confirm in your own account with a query on the metering history view.
Where Adapters fits
Adapters maps source objects onto typed Snowflake tables and merges changes on the source id, with flat monthly plans from $49 and no instance hours to count. It is a different trade, not a Snowflake-managed connector, so it suits teams that want one predictable bill. See the Snowflake ETL tools comparison and pricing.
Load Snowflake from your apps on a price you can forecast
Map the objects once, land exact types, then sync only what changed. Plans from $49 a month, with no instance hours to count.
The live demo needs no card, and Starter is $49 a month.