Snowflake cost and savings calculator

Introduction

If you want to gain a better understanding of Snowflake pricing works you are in the right place. On this page we explain how Snowflake usage-based pricing model works. You can estimate your potential Snowflake cost, as well as calculate how much you can potentially save by using IOMETE’s flat fee pricing (spoiler alert: > 50%).

About Snowflake

Snowflake (NYSE: SNOW) provides a data warehouse-as-a-service (DWaaS) platform. This means that Snowflake provides all the necessary hardware, software, and infrastructure for a data warehouse. Under the hood Snowflake leverages the cloud infrastructure of AWS, Microsoft Azure or Google Cloud.

Snowflake was founded in 2012 by Benoit Dageville, Thierry Cruanes, and Marcin Zukowski. The company was initially funded by venture capital firms Sutter Hill Ventures, Altimeter Capital and Sequoia. Snowflake went public in September 2020 and was one of the largest IPOs in recent years. Snowflake is headquartered in Bozeman, Montana, and has offices in North America, Europe, and Asia. Snowflake has 6,300 employees and is growing rapidly. Snowflake’s Annual Recurring Revenue in 2022 was $2.5Bn and its market capitalization around $50Bn at the moment of writing (August 2023).

About IOMETE

IOMETE is an early-stage startup backed by Y Combinator. We started working on IOMETE three years ago and built a modern and open data lakehouse that outperforms incumbent legacy solutions.

Usage-based Pricing

Snowflake pushes a “pay-as-you-go” or “usage-based” pricing model, where the more data one computes, the more one pays. The formula is simple: (a = b x c):

(a) Snowflake compute cost = (b) Number of Compute Credits X (c) Price per Compute Credit.

Snowflake positions the usage-based pricing model as a customer benefit with slogans like “you-only-pay-for-what-you-use”. As data tends to grow over time, “you-only-pay-for-what-you-use” can quickly become “you-pay-for-everything-you-use”…

This is reflected by Snowflake’s net retention rate. Snowflake's net retention rate is the percentage of revenue that the company retains from its existing customers in a given period. Snowflake's net retention rate has been consistently high in recent years. In 2021, the company's net retention rate was 151%. In 2022, the company's net retention rate was 142%. So if the average customer paid $100 in 2020, that would be 51% higher cost for that customer - or $151 - in 2021. The average customer paid $100 in 2021 and paid $142 in 2022, a 42% cost increase.

(source: Snowflake SEC filings).

Snowflake Compute Credit

Snowflake is a cloud-based data warehouse. It leverages AWS, Microsoft Azure or Google Cloud to run compute, but Snowflake does not disclose what specific AWS, Azure of GC nodes it is using. We wrote a dedicated post What is a Snowflake Compute Credit that we recommend reading. Snowflake Compute Credits are priced between $2 and $4 depending on the cloud (i.e. AWS, Azure, GC) and the cloud region. Snowflake does not disclose which nodes it is running under the hood. For example, if you select a XS node, you have no idea what node Snowflake is running under the hood. It’s like buying a bag of coffee for $2, but you don’t know if you’re getting 300 or 400 grams or whether you get Robusta or Arabia beans.

Usage-based vs flat fee pricing

Usage-based pricing was popularized by AWS. It makes sense that AWS applies variable pricing because their cost structure is also mostly variable. The more you compute, the more AWS needs to invest in servers, buildings, infrastructure. However, this is not necessarily true for Snowflake (or IOMETE for that matter). Snowflake is software that leverages the hardware of AWS/Azure/GC. The distribution cost of software is zero. The incremental cost of running one more instance is close to zero for Snowflake, so why a usage-based pricing model? We can only guess. Yes, it sounds logical, you only-pay-for-what-you-use. Example: But what about large organizations that could have purchased a node for $0.50 from AWS and pay $3 per compute credit to Snowflake? They finance that $2.50 mark-up forever. There is a point at which flat fee pricing becomes much cheaper than usage-based pricing. Could it that usage-based pricing is pushed because it is most beneficial to Snowflake?

A different pricing philosophy

IOMETE charges a flat fee for access to our platform. On top of that the customer pays the AWS (or Azure or GC) node without any mark-up (directly to AWS/Azure/GC with no interference from IOMETE). This has a bunch of advantages:

1. The customer can save a lot of money (at least 50% in most cases).

1. The customer knows exactly what node(s) it is using.

1. The customer can optimize its compute spend on AWS, Azure and GC by using reserved and spot instances.

1. The data remains in the customer’s perimeter. The customer does not get locked-in.

Perverse incentive vs aligned interests

Flat fee pricing puts IOMETE in a position where our interests are aligned with those of you, the customer. Why? We have no interest in selling you more compute credits. Compute should not be a goal in itself. In stead we align our interest with yours. You pay us a platform fee. We help you reduce compute spend on AWS, Azure, GC.

Why do we charge a flat fee? Don’t you want to become as rich as Snowflake?

Well it does not cost us anything extra whether you are computing 10 hours per week or 10,000 hours per week. It is software after all. Snowflake has done extremely well for itself and its investors, but they have made themselves vulnerable to disruption. Snowflake customers can save a ton of money by switching to IOMETE. The more compute credits you are using, the bigger the savings!

We are a startup with a different pricing philosophy. We don’t need to satisfy the stock market each quarter. And: Each $1 of new revenue is $1 we didn’t have before.