Snowflake bills can shock even the most experienced data teams. Costs can jump from six figures to seven in a single quarter – without warning.
The culprit? Runaway compute costs and inefficient query patterns. The key to keeping that bill from shocking you is to know what goes on behind the Snowflake pricing scene and how to optimize your setup.
This guide will demystify Snowflake costs. We’ll walk you through:
- How Snowflake pricing works
- What influences Snowflake prices
- How to optimize Snowflake costs for a better bill
- Snowflake FAQs
How Does Snowflake Pricing Work
Snowflake charges you based on usage per second. There are only three services that Snowflake will charge you for:
- Data storage
- Compute
- Cloud services
These three services align with Snowflake’s three major architecture layers, data storage, data warehouses (compute), and data transfer (cloud services).
Still with us?
You should also know you don’t technically pay cash for these services. You pay in Snowflake credits.
Let’s start there first.
Snowflake Credits
Think of Snowflake credits as arcade tokens. You’re not paying dollars directly, but instead buying credits up front. You then “spend” those credits on compute resources. A single credit roughly equals a single hour of compute time for an X-Small warehouse.
Each Snowflake layer uses credits in a different way though. How many credits you’re charged also depends on your edition (Standard, Enterprise, or Business-Critical) and your account type (Pre-Purchase Capacity or On-Demand). But we’ll get into all of that in a bit.
Data Storage Costs
You’re charged for data storage based on the daily amount of average data bytes stored. This is all data, including compressed and uncompressed files, historical data, and database data.
You’ll be charged a monthly rate per terabyte, though that per TB number varies depending on your account type and plan.
Snowflake is on your side when calculating that final bill. It will automatically compress and optimize your table data and bill you based on that total.
Virtual Warehouse Costs
A virtual warehouse is where your queries, loading data, and other DML operations work. Snowflake credit costs vary depending on:
- Warehouse size
- How many warehouses you have
- How long the warehouse is running
Something important to remember here: you’re paying for services per a second, but Snowflake has a minimum charge of 60 seconds when starting. That means even if you run your warehouse for ten seconds, you’re still going to pay for the full minute.
You can pick from ten different sized warehouses. The larger the warehouse, the faster it will work and the more data it can process, but the higher the cost.
It can be tempting to immediately jump up to a large warehouse, thinking that will save you money, but that’s not the best way into warehouse optimization. If you don’t have a third-party tool to automate that for you like Yuki, you’re looking at a trial and error process to optimize your spending.
Cloud Services Costs
Snowflake cloud services are the bones behind data storage and compute functions (e.g., SQL lAPI, query parsing, metadata and infrastructure management, etc.). These services are all powered by compute resources which, you guessed it, you pay for with Snowflake credits.
The key difference when it comes to cloud services is that Snowflake only bills for services that exceed 10% of your daily compute resource usage.
Here’s how it works:
If you use 100 compute credits in a day, Snowflake gives you 10 cloud service credits for free (10% of 100). If you only used 8 cloud service credits, you pay nothing extra. If you use 15 cloud service credits, you only pay for 5 (the amount over your free allowance).
If your cloud services end up being less than 10% of your daily compute, Snowflake will just charge you based on how many cloud service credits you’ve used for that day.
What Influences Snowflake Prices
Snowflake prices fluctuate depending on your:
- Snowflake plan
- Account type
- Region
- Platform
- Hidden costs
Let’s take a closer look at how each of these variables leads to a higher (or lower) monthly bill.
Snowflake Plans
Snowflake has four different plans you can choose from: Standard, Enterprise, Business Critical, and Virtual Private Snowflake (VPS).
Let’s break each of these down:
Standard
This plan includes:
- Complete SQL data warehouse
- Secure data storing
- Premier support
- Federated authentication
- External functions
- snowsight
- One day of time travel
- Data Marketplace access
- Enterprise-grade encryption
Enterprise
In addition to all features included in the Standard plan, the Enterprise plan includes:
- Multi-cluster warehouse
- Annual rekeying of all encrypted data
- Materialized views
- Search Optimization Service
- External data tokenization
- Up to 90 days of time travel
Business Critical
Business Critical includes these additional features:
- HIPAA support
- PCI compliance
- AWS, Azure, and GCP Private Link and Private Service Connect support
- Business continuity and database failover and failback
- AWS API Gateway Private Endpoints support
Virtual Private Snowflake (VPS)
In addition to all of the features listed above, VPS includes:
- Customer-dedicated metadata store
- Customer-dedicated virtual servers
Account Type
There are two different account types you can pick from: Pre-Purchase Capacity and On-Demand. If you’ve ever worked with a cloud system before, you’re already somewhat familiar with these plans.
On-Demand gives you the flexibility to purchase as you go, but you’ll have to pay more. Storage, for example, will cost you $40 per TB per month if you’re using AWS in the US East Region.
Pre-Purchase is better for organizations where you have a clearer idea of how much support they’ll need. You are committing to spending and using a set amount, but you’ll save a fair bit, especially if you’re looking at large purchases. If you were to buy that same service with a Pre-Purchase cost, you’d only have to pay $23 per TB per month.
Region
Depending on where your Snowflake workload is based, you may have to pay a bit more – or less. Costs vary from country to country and region to region.
You’re looking at paying as low as $90 per TB for data transfer to a different cloud if you’re based in the US East (Ohio) region, or as high as $150 if you’re located in South America East 1 (Sao Paulo).
Platform
Snowflake works natively on AWS, GCP, and Azure. Each of those platforms offers a slightly different price for Snowflake features.
Here’s how those prices vary if you were looking at data transfers:
- AWS US East (Northern Virginia) Region: $90 per TB
- Google Cloud (US East 4 – N. Virginia) costs $120-$190 per TB
- Microsoft Azure (East US 2 – Virginia) costs $87.50 per TB
Hidden Costs
These pesky features are often the culprit behind your suddenly spiking bill. Keep an eye out for:
- Time travel storage: Ask yourself how much historical data you really need, and if it’s worth extending up to 90 days for the amount you’re paying for storage.
- Clustering keys: This can help with query performance, but at the cost of additional compute resources.
- Automatic clustering: Continuously reorganizing data helps improve optimal performance, but watch out for how many credits Snowflake consumes in the background.
How to Optimize Snowflake Costs
The first step to improving your Snowflake costs is understanding what goes into your bill. Here’s what you should do next:
- Usage monitoring
- Optimize your warehouse spend
- Review your queries
- Try out Snowflake Optimization Services
- Consider automated tools
Usage Monitoring
Before you start optimizing, you need a baseline. Snowflake has built-in usage monitoring tools that show exactly where your credits are going. Look out for:
- Warehouses running longer than expected
- Peak usage times that need a little auto-scaling
- Large warehouses used for small queries
Optimize Your Warehouse Spend
Like we mentioned above, finding the perfect warehouse is a bit of a Goldilocks situation. Too small, and you’ll be left with slow queries and processing speed. Too fast, and you’re paying for power you don’t need.
If you’re A/B testing your way to the perfect warehouse, you’re going to want to start small first and work your way up. Without a third-party automation tool like Yuki, you’ll have to do this manually, but it can be worth the effort if you have the time.
Make sure you’re using the warehouse auto-suspend feature. This is your biggest cost-savings opportunity. You can have your warehouses automatically shut down after periods of inactivity (the default is 10 minutes, but you can set this even lower).
Review Your Queries
Did you know your SQL code is likely inflating your Snowflake costs?
Carefully review your queries. Improve your Snowflake query optimization skills by looking for ways you can:
- Improve your filter use
- Avoid unnecessary columns and joins
- Use Snowflake features like materialized views or clustering keys
The more efficient your query, the less compute power Snowflake has to use – and the lower your final bill.
Try Out Search Optimization Service
Snowflake Search Optimization Services (SOS) is an Enterprise Edition (or higher) feature that uses an advanced behind the scenes indexing structure. These indexes make it easier for you to find specific values, reducing query latency and resource usage for certain queries.
Consider Automated Optimization Tools
Manual adjustments for Snowflake cost optimization only get you so far. Plug-and-play tools can automatically right-size your warehouses, improve your queries, and prevent runaway costs – all without a developer.
Our customers start saving an average of 37.6% when they use Yuki. Get a free analysis so you can find out how much you can save on your Snowflake costs today.
