Amazon Web Services (AWS) launched its service in 2006. Organizations could rent virtual computers and storage in the cloud to store their data and run their software applications. When work began in 2012 to build Snowflake’s cloud data warehouse, it was obvious to our founders that Snowflake should run on top of AWS as our first cloud infrastructure partner.
The engineers were working hard to develop our product when the announcement came. AWS launched Redshift—their own data warehouse for the cloud. We were still three years from GA, or becoming “generally available.”
Worse, AWS charged rock-bottom prices for Redshift. It was practically included with the price they charged for their computing-as-a-service, which was the real cost we would have to pass onto our customers for eventually running Snowflake on AWS. This was not a new tactic in the tech industry. Long before the cloud was commercially available, behemoth hardware companies such as IBM created their own databases and structured their pricing to make it look like their software was free to customers. Not long after, in the 1990s, a series of application software companies became the dominant players in their respective sectors within tech. Many of them had solutions that didn’t work well but they had incredible influence over customers and deep pockets to market in ways that obscured their product deficiencies. Either way, these tech vendors had established a history of getting customers to adopt substandard technology in order to close a deal.
AWS was already the dominant player in cloud computing and storage. Soon after, they took aim at the software stack, offering a data warehouse and additional products. With Redshift commercially available, everyone—including the investor community—thought AWS was going to eat our lunch. “Why would we invest in Snowflake only to have AWS kick the crap out of you?” That was their attitude. Yet, the reality of AWS crushing Snowflake before our product was still in development jarred us. Snowflake Co-Founder Benoit Dageville, who is a world-class database architect, heavily considered the option of packing up and going home. He wouldn’t have been the first tech visionary to realize that even the best product from a startup couldn’t survive whatever came from a behemoth of a competitor.
What AWS revealed about the cloud data market
But then two things emerged that transformed Redshift from being our biggest competitor to the product that helped jumpstart Snowflake’s commercial success. AWS signed more than 1,000 Redshift customers in just nine months. Interestingly, AWS did not acquire these customers from the existing on-premises data warehouse market. They were organizations choosing the cloud, which offered theoretically unlimited amounts of computing and storage. There were smaller organizations that didn’t already have acres of their own on-premises hardware configurations. They didn’t have an aircraft carrier to turn around.
AWS revealed there was a market for cloud data warehousing. It also made a serious mistake. In its rush to offer a cloud data warehouse, it adapted ParAccel, a database designed for on-premises computing environments, called it Redshift, and offered it in the cloud.
When the business world started to realize that the cloud could be as safe as their own data centers—or even safer—software companies made haste to produce cloud versions of their existing solutions. Many of them also rushed that process, understandably so. Simply porting an on-premises solution had its benefits. Customers could shutter their own data centers and simply access those same solutions via the Internet. But the benefits of a ported solution didn’t extend much beyond that. Those solutions failed to take advantage of the enormous resources cloud computing offered. A solution built for the cloud, such as Snowflake, could do that.
Fortunately, Snowflake’s engineers continued to develop our data warehouse for the next two to three years. Redshift’s appeal grew immensely. In addition to seeding customers that were cloud-first, AWS started to disrupt the traditional data warehouse market. It was much less expensive than a legacy data warehouse, so customers used Redshift alongside those legacy solutions. It’s rare that a large customer will execute a wholesale dump of an incumbent product. Instead, they will buy a competing product for a few specific features. It’s costly and cumbersome but it can start a much-needed look at a staged migration of all its computing solutions to the cloud.
It was June 2015 when Snowflake’s cloud-built data warehouse became commercially available. It still lacked many staple features that customers raised with us during the product’s development. But that didn’t stop the sales team from pounding the phones and the pavement. AWS Redshift was still winning tons of accounts. We were competing against their sales reps and losing most of the time. These accounts had deployed Redshift only recently and the product was working for them. Also, even if a customer thought Redshift and Snowflake were a tie, or Snowflake had an edge over Redshift, they would not choose us. We had fewer than 100 employees. AWS had been in business for more than 10 years, had thousands working for them, and was the anointed tech child of Amazon. Customers often choose stability over a startup.
‘The best thing that ever happened to us’
Bob Muglia, Snowflake’s CEO at the time, was on the East Coast, attending sales calls with Vince Trotta, who turned to Bob and said: “You know, I think these Redshift accounts are the best thing that ever happened to us. All of these accounts are gonna come around 18 months from now, when they run out of gas on Redshift and we’re going to win them over.”
That’s exactly what started to happen. Redshift customers demanded more of the product, which was still limited by its legacy architecture designed for on-premises environments. “They did a classic (Jeff) Bezos deal,” Bob once said. “Amazon didn’t buy the company. They bought the codebase. And they took the database, called it Redshift, stuffed it in the cloud, and priced it dirt cheap. It reinvigorated the data warehouse market because the product actually worked until you hit the wall, and then it was a terrible product.”
Redshift couldn’t scale to handle ever-increasing workloads that required more data and more computing. The AWS computing power was there but its own data warehouse couldn’t exploit that resource beyond what it was designed to do inside customers’ own data centers. We started to win those accounts because Redshift couldn’t scale. It was the perfect competitor for us. A behemoth and well-respected technology provider that was first to market with great technology for on-demand storage and computing in the cloud. Then it offered a data warehouse on top of that foundation but one that would eventually run out of gas. Later on, at a Snowflake sales kickoff event that included about 30 of us, Bob stepped onto the two-foot-high stage, sat down on its edge, held the cordless mic and uttered in a tone lower than normal from him and said: “Thank God for Redshift.”
Excerpted with permission from the publisher, Wiley, from Make It Snow: From Zero to Billions: How Snowflake Scaled its Go-to-Market Organization by Denise Persson and Chris Degnan. Copyright © 2026 by John Wiley & Sons, Inc. All rights reserved. This book is available wherever books and eBooks are sold.