Over the past couple of years, companies ranging from small startups to large enterprises have shown a growing interest in public clouds such as AWS. So many AWS case studies are not surprising, given that both the operation expenses model of public clouds and their abundance of available managed services make it very easy to start a business in the cloud or migrate and existing one to it.
While we did have an idea of the scale at which some businesses were utilizing AWS, some of the recent headlines still caught many by surprise. “Pinterest has spent $309 million on Amazon’s cloud since 2017 as part of a $750 million contract,” one read. “Slack to spend at least $250 million on Amazon Web Services over five years,” and “Lyft committed to spend $300 million on AWS between 2019 and 2021,” are others. This article will focus on why these major companies are paying such large sums to use AWS infrastructure, and why they are making such huge commitments to public cloud use.
AWS Case Studies: Lyft, Slack,and Pinterest
Let’s start by taking a closer look at how each of these companies uses the AWS cloud as well as what their current contractual obligations with Amazon are.
After filing for an IPO, Lyft revealed that it will have to spend at least $300 million (about $8 million per month) on AWS cloud services before the end of 2021. Apparently, the contract states that if Lyft doesn’t hit the $300 million mark, they will have to pay the difference.
When Lyft first started back in 2012, they ran their workloads in the AWS cloud using only a couple of EC2 instances. Now, with their service facilitating around 14 million rides each month, Lyft has a significantly larger infrastructure. This rapid growth was made possible by AWS’ on-demand elasticity. This same elasticity allowed Lyft not only to scale up quickly (up to 8 times during peak periods), but also to scale back down quickly in order to avoid unnecessary overspending. The Lyft cloud environment depends on many other AWS services, including ECS for their containerized microservice architecture, Kinesis and DynamoDB for storing GPS coordinates, and Redshift for acquiring customer insights. They also rely heavily on Spot instances for their Continuous Integration (CI) system, which saves them 70-90% on compute resources.
Slack, too, has significant commitments to the Amazon cloud and are just one of the many AWS case studies that have invested a signifcant sum for a period of time—somewhere in the vicinity of $250 million over the next five years. This means that about 10% of their annual revenue is being spent on processing and delivery infrastructure.
Slack has been running on AWS since its inception in mid-2013. Slack’s predecessor, a company called Tiny Speck, used AWS back in 2009 when the public cloud was still in its infancy. Today, Slack is being used by numerous companies, and the number of messages flowing through the app every day is in the millions. The company is only getting larger, thanks to its 15% weekly user growth rate.
Slack relies on simple cloud architecture (EC2 instances behind Elastic Load Balancers, S3, etc.), but it benefits greatly from the resource availability that AWS can provide. While adding capacity in traditional data centers took weeks, it is a non-issue with AWS. For a fast growing company like Slack, the cloud was not only the best choice, it was probably the only one. In addition, the managed nature of most of AWS’ services has allowed Slack to minimize their IT management involvement in order to focus on their much sought-after product instead.
Pinterest has also recently filed for an IPO, allowing the public the opportunity to look into their contracts with Amazon’s public cloud service. According to their agreement, Pinterest will spend $750 million on AWS by 2023. They have already spent over $300 million since 2017, so, by the looks of it, they will have no problem spending the rest by the end of the contract’s term. As with Lyft and Slack, Pinterest will have to pay any difference if their AWS use falls short.
Pinterest also started on AWS back in 2010. Their business has since grown to a staggering 200 million active monthly users. In 2015, Pinterest was already looking into employing a containerized solution. They first moved to Docker containers, then they started using self-hosted Kubernetes for orchestration. This shift allowed them to optimize their cloud infrastructure, simplify their deployment and management processes, and save on infrastructure costs. In addition to using AWS for processing, Pinterest also uses various other services for storage and data analysis, allowing them to put more focus on their product.
What Can Be Learned Here?
When startups begin to grow exponentially, the cloud seems to be an easy (if expensive) solution that provides new companies with the agility they need as well as the ability to scale almost infinitely very quickly. When companies reach these high levels of hardware requirements, owning and maintaining the necessary bare metal is typically not a viable solution. For all three of these companies, the advantages that AWS cloud provides seem to outweigh its enormous costs.
Still, why such do these AWS case studies have such a huge commitment?
Well, it actually makes sense. While we don’t know the exact details of these companies’ contracts with AWS, we can assume the commitments result in long-term cost savings and capacity reservations. AWS already has services like Reserved Instances, which provide companies with huge discounts (up to 75%) for long-term commitments. On top of that, they also provide guaranteed capacity.
These particular AWS case studies have certain leverage as well. Since Lyft, Slack, and Pinterest are huge spenders, they can negotiate better terms with Amazon and secure significant discounts for Amazon’s cloud resources. Given the monthly bills these companies are paying, any negotiated discount makes a big difference.
One of the main reasons for running on AWS has always been its availability of resources. Companies like Lyft, Slack, and Pinterest simply can’t afford to be left without desired capacity. In order to keep growing, their requirements must be met, and these contracts guarantee them the ability to do so.
What If These Companies Wanted to Opt Out of AWS?
Even without their contractual obligations, these companies would face a daunting task if they decided to move out of AWS. The re-architecting of components would require an enormous amount of work. These companies are clearly not interested in focusing on infrastructure, so it is safe to say that they will avoid such a move at all costs. Migrating data out of AWS would also be so costly that no company is likely to choose that option.
Of course a move like this is possible; Dropbox recently dropped Amazon for their own infrastructure. Still, not many companies are willing to undertake a multiple-year project of that size.
Currently, Lyft, Slack, and Pinterest are fully invested in AWS and will continue to be for at least a couple more years. While AWS provides them with a number of important benefits, it is hard to know whether or not committing to AWS was the best move for them in the long term. These AWS case studies will be interesting to follow. Will they continue to rely on Amazon for cloud resources in the future? Or will they look back at this commitment as a huge mistake while trying to extract themselves from massive vendor lock-in? Only time will tell.