Snowflake in Power BI: Should You Use DirectQuery or Import?

Part A - Understanding Snowflake Pricing Model

Snowflake uses a pay-as-you-go pricing model based on consumption of credits, which are charged for:

1. Compute usage (via Virtual Warehouses)

2. Cloud services (minimal unless metadata load is very high)

3. Storage (separately billed)

For Power BI use cases, Virtual Warehouse compute cost is the main area of focus.

Let's first quickly talk about Storage and Cloud Services

Cloud Services Cost (Minimal)

This is charged in addition to compute but is usually a small % of your total.

It covers:

• Query parsing and optimization

• Transaction coordination

• Metadata services (SHOW commands, etc.)

• Authentication and security operations

  
  

It’s not charged per warehouse, but per account, and only when it exceeds 10% of total compute cost. So it’s typically negligible for normal users.

Storage Cost (Monthly, pay-per-usage)

This is charged based on the volume of data stored, not credits.

You’re billed per terabyte per month — typically ~$23/TB/month (varies by region/cloud provider)

Types of storage:

• Active storage – your tables, views, staging areas

• Time Travel storage – snapshots of data (if enabled)

• Fail-safe – 7-day read-only retention, not user-configurable

Storage is NOT tied to warehouse size — it's flat-rate pricing.

There are some other costs if you use more features, but we won’t get into it now.

Compute Cost (Virtual Warehouses) – the main one

This is what most people think of when they say “Snowflake costs money.”

• Every query you run (manually, through Power BI, dbt, etc.) runs on a virtual warehouse

• Warehouses come in sizes: X-Small, Small, Medium, Large...

• Each size consumes credits per hour

• The default credit cost is $3 per credit (unless you have custom/discounted pricing)

Power BI DirectQuery consumes from here for every click.

Import only consumes from here during scheduled refreshes or Loading Data in Development stages.

Summary Table

Layer	What It Covers	Billed As	Approx Cost
Compute	Query execution (e.g. Power BI)	Credits/hour	$3/credit
Cloud Services	Metadata, auth, query planning	Credits (if >10%)	Usually negligible
Storage	Data at rest (including time travel)	$/TB/month	~$23/TB
Optional Extras	Clustering, Snowpipe, Tasks	Credits or per event	Varies

Understanding Compute Cost in More Detail

What is a “Credit” in Snowflake?

A credit is Snowflake’s basic billing unit for compute. You don’t pay per query, per second, or per user — you pay based on how long a warehouse runs, and how large that warehouse is.

Credit Usage = Warehouse Size × Time

Each warehouse size consumes a different number of credits per hour:

Please Note - The below table values might change. Please refer to your actual agreement with Snowflake for correct calculations.

Some places the cost for X-Small shows 0.5 credits per hour and some places show 1 credit per hour. Please use your actual bills for better Judgement.

We have taken 1 credit per hour as per the Official Documentation of Snowflake Pricing.

Warehouse Size	Hours	Credits per Hour	$ Cost (at $3/credit)
X-Small	1	1	$3.00/hour
Small	1	2	$6.00/hour
Medium	1	4	$12.00/hour
Large	1	8	$24.00/hour

You only pay for the warehouse while it's running — and Snowflake bills per second, with a 60-second minimum.

Example 1 – Cost of a 5-Minute Query

Here’s a breakdown of the cost of running a single 5-minute query across all main Snowflake warehouse sizes, assuming standard pricing of $3 per credit.

Warehouse Size	Credits per Hour	Time (hrs)	Credits Used	Cost per Credit	Total Cost
X-Small	1	0.0833	0.0833	$3	$0.25
Small	2	0.0833	0.1666	$3	$0.50
Medium	4	0.0833	0.3332	$3	$1.00
Large	8	0.0833	0.6664	$3	$2.00

Summary:

• X-Small = Cheapest, but slowest

• Small = Best balance for small workloads

• Medium+ = Fast, but expensive for just one query

• Large and X-Large = Overkill unless handling many concurrent users or heavy joins

Part B - Understanding Snowflake Pricing Model in context to Power Bi

Let’s start with the obvious — Snowflake is powerful, fast, and cloud-native. But when it comes to integrating Snowflake with Power BI, many people fall into one of two extremes:

• Option 1: DirectQuery — real-time queries on live Snowflake data

• Option 2: Import — snapshot the data once and refresh it periodically

  
  

So which one should you actually use?

Let’s go deeper.

DirectQuery vs Import Mode – What’s the Difference?

DirectQuery means every time a user clicks a visual in Power BI, a live SQL query is sent to Snowflake.

Import means the data is loaded from Snowflake into Power BI ahead of time. Visuals work off that snapshot — fast and local.

Here’s a side-by-side view for client education:

Feature	DirectQuery	Import
Speed for user	Depends on warehouse size; usually slower	Very fast (local cache)
Cost impact	High — each click = Snowflake query	Minimal — only during refresh
Max flexibility	Always live data	Data refresh needed
Power BI functionality	Limited (some DAX/PQ not supported)	Full functionality
Performance tuning	Depends on warehouse + modeling	Modeling only

Now let’s apply this to a real-world scenario.

Scenario Assumptions

Direct Query -

• 500 end users

• Each runs 5 queries per day

• Each query takes ~3 seconds

That’s roughly 63 hours of Snowflake runtime per month if using DirectQuery.

Using DirectQuery (63 hrs/month)

Warehouse Size	Monthly Runtime	Monthly Cost
X-Small	63 hrs	$189.00
Small	63 hrs	$378.00
Medium	63 hrs	$756.00
Large	63 hrs	$1512.00

Import -

• Lets assume it takes 1 minute for the refresh to run

• You have switched on Schedule Refrsh for once per day

• It means 30 minutes per month or 0.5 hrs

Using Import Mode (0.5 hrs/month)

Warehouse Size	Monthly Runtime	Monthly Cost
X-Small	0.5 hrs	$1.50
Small	0.5 hrs	$3.00
Medium	0.5 hrs	$6.00
Large	0.5hrs	$12.00

Cost Comparison (Snowflake Pricing)

Warehouse Size	DirectQuery Cost	Import Mode Cost
X-Small	$189.00	$1.50
Small	$378.00	$3.00
Medium	$756.00	$6.00
Large	$1512.00	$12.00

So... Use Import Mode?

Yes — in most cases, Import mode is the obvious winner:

• Huge cost savings

• Faster user experience

• More Power BI features available - Like Power Query, Advance Dax, Dynamic Filter functions

But there are things to watch for.

⚠️ When Import Might Cause You Problems

• Your dataset is too large to fit in memory

• You need true real-time data (stock prices, logistics feeds)

• You’re handling event-based streaming data that changes every few seconds

  
  

If any of these apply, then DirectQuery with a Medium or Large warehouse may be justified.

Please Note – Power BI also compresses data.

It uses an in-memory engine called VertiPaq, which reduces file size through columnar storage, dictionary encoding, and smart data typing. It also applies query folding, auto relationships, and caching to boost performance. We won’t get into more detail on this here. 

These optimizations make Import mode highly efficient for analytical workloads.


In conclusion, your 1 GB of data in Excel or Snowflake might shrink to just a few hundred MB once loaded into Power BI.

Our Recommendation

Start with Import mode.

Test performance.

Only consider DirectQuery if you know:

• You need real-time data, and

• You’re okay with higher Snowflake compute charges

For 95% of reporting scenarios, Import Mode gives better control, better speed, and massively lower Snowflake costs.

If your Snowflake bill is growing — and you’re using Power BI DirectQuery — now you know why.

Let’s talk about optimizing it together.