How to Connect Snowflake to Power BI

Power BI has a native Snowflake connector that supports both Import and DirectQuery, so connecting them is straightforward; the decisions that matter are which storage mode to use, how to handle the gateway and authentication, and how to stop Snowflake compute costs from creeping up. Import gives fast, stable reports refreshed on a schedule. DirectQuery gives live data and handles very large datasets, but pushes queries to Snowflake on every interaction, which costs compute. This guide walks through both, plus the gateway and cost-control practices that keep a deployment healthy.

Import versus DirectQuery

Import mode loads a copy of the data into Power BI's in-memory engine. Visuals are fast, DAX is unconstrained, and Snowflake is only queried at refresh time. This is the default recommendation for most reports. The limits are dataset size and data freshness, since users see data as of the last refresh.

DirectQuery leaves data in Snowflake and sends a query for each visual interaction. You get near-live data and can report over datasets too large to import, but visuals are slower, some DAX is restricted, and every slice, filter and page load can spin up Snowflake compute. Composite models let you mix the two, importing small dimension tables while keeping a huge fact table in DirectQuery.

For most organisations, Import with a sensible refresh schedule is cheaper and faster. Reserve DirectQuery for genuine real-time needs or volumes that cannot be imported.

Gateways and authentication

When you refresh Import datasets on a schedule or run DirectQuery from the Power BI Service, you generally need a data gateway unless the connection is a fully supported cloud-to-cloud path. For production:

• Size the gateway for concurrent refreshes and queries, and use a gateway cluster for high availability rather than a single machine.
• Use Microsoft Entra ID single sign-on where possible so Snowflake row-level security and roles apply per user, rather than everyone sharing one service account.
• Manage credentials centrally and rotate them; avoid embedding personal logins in shared datasets.

Getting the gateway and identity model right early avoids painful rework once reports are in use across the organisation.

Controlling Snowflake costs

Snowflake bills for compute, so a careless Power BI deployment can quietly run up a large bill, especially with DirectQuery. Practical controls:

• Prefer Import unless freshness genuinely requires DirectQuery. A scheduled refresh costs far less than thousands of interactive queries.
• Right-size and auto-suspend warehouses. Use a dedicated, small warehouse for BI workloads that suspends quickly when idle.
• Aggregate in Snowflake. Build summary tables or views so Power BI reads pre-aggregated data instead of scanning raw fact tables.
• Use Power BI aggregations and composite models so detail queries only reach Snowflake when a user drills down.
• Limit refresh frequency to what the business actually needs; hourly refresh on data reviewed once a day is waste.
• Monitor query history in Snowflake to find expensive, frequently run report queries and optimise them.

These steps routinely cut BI-driven Snowflake spend without users noticing any loss of capability.

Modelling for performance

Whichever mode you choose, a clean star schema beats a tangle of wide tables. Push joins and heavy transformations into Snowflake views so Power BI consumes a tidy, query-friendly shape. Keep DirectQuery models narrow, index-friendly and free of complex calculated columns. Document the metrics layer so report authors reuse measures rather than reinventing them.

Getting it done well

A well-architected Snowflake-to-Power BI deployment balances speed, freshness and cost rather than defaulting to one mode for everything. Our Power BI consultants in the UK design the storage mode, gateway topology and cost controls together, and our data integration and warehousing work covers the pipelines that keep Snowflake feeding Power BI. A typical engagement starts with a fixed-scope Trusted Numbers Review (£2,000–£4,000 as of 2026, confirm current pricing). If your data lives in the wider Microsoft stack, see our Power BI and Microsoft Fabric work, and our guide on connecting Dynamics 365 to Power BI.

Connect Snowflake to Power BI in a way that stays fast and keeps compute spend predictable. Book a Trusted Numbers Review and we will review your storage mode, gateways and Snowflake costs, then map the improvements that matter most.