It’s no surprise that Amazon Web Services (AWS) powers many industry-leading companies. It provides on-demand access to infrastructure and frees up time to focus on feature development and product growth. Gone are the days of manually replacing broken servers and upgrading machines yourself; hello, increased security, automatic server maintenance, and lower costs.

Yet these costs come with caveats, a16z’s Sarah Wang and Martin Casado write extensively about in their report on the “trillion-dollar paradox” of cloud. Most saliently, Wang and Casado point out the classic paradox in which the cloud delivers on its promise early in a company’s journey, but that “the pressure it puts on margins can start to outweigh the benefits, as a company scales and growth slows.”

A16z estimated that for large-scale operations, cloud costs can double a company's infrastructure bill. The analysis contextualized the magnitude of cloud spend as a percentage of the total cost of revenue (COR), and estimated that Palantir's cloud spend made up 38% of its COR, and Asana's cloud spend made up 63%, for example.

Dropbox saved a total of $75m by moving off of the cloud in an infrastructure optimization overhaul that took place in the two years leading up to their IPO. Gross margins increased from 33% to 67% between 2015 and 2017, which the company primarily credited to infrastructure optimization and an increase in revenue.

Mature companies eventually will have to contend with infrastructure choices made in earlier, high-growth stages. As a16z points out of the cloud paradox, not having to worry about infrastructure management to focus on feature development is a blessing, though optimization becomes higher-priority with maturity. As it was in the case of Dropbox, meaningfully cost-optimizing the same infrastructure that enabled the company’s initial success can take years later on down the line.

While companies like Dropbox, CrowdStrike, and Zscaler explored cost-optimization options that involved moving off of AWS and back to on-premise data centers, it is possible to save as much as 57% through optimizing AWS spend with the likes of AWS Savings Plans and Reserved Instances.

These discounts are underutilized by most companies because of cloud pricing, which has become so complex to the point that there are companies with multi-billion dollar valuations, like Turbonomic’s (acquired for $2B by IBM), born out of the problem of the complexity of cloud management.

Conventionally, CIOs at the helm of an organization’s IT spending could advantage of these discounts through close and careful collaboration with finance and engineering departments. Organizations have to forecast money to commit upfront, growth trajectory, and compute resources for the next three years, by months and quarters. If an organization can pull off the dance of coordinating between finance and engineering, it is possible to achieve 50% savings every year on its AWS bill.

For CIOs to engage with business partners more effectively, they must "put technology aside and focus on their internal and external customers," as Miriam McLemore, Enterprise Strategist at AWS writes.

Additionally, there are billions of different pricing scenarios for customers, so finding the right balance within a given billing mode’s options, across at least 200 instances, between Reserved Instances and Savings Plan, is no trivial task.

Making the most of your AWS savings can be a daunting task, but with a little bit of know-how, you can get started on the right foot. One of the best ways to save on AWS is to optimize your workloads. Take a look at your current applications and see where you can reduce the number of servers you’re using, or where you can use more cost-effective instances.

For example, spot instances can be a great way to save money on compute power, and hybrid and burstable instances can help you save on storage costs.

Another key area to focus on is your data. Reducing the amount of data you store can go a long way towards reducing your AWS bill. You can also optimize your data transfer and storage costs by using Amazon S3 Glacier, which offers extremely low-cost storage for infrequently accessed data. Further, using reserved instances can be a great way to save on AWS. By committing to a certain instance type for a year or more, businesses can get discounted rates on instances.

Finally, make sure you are taking advantage of all the available tools and services that AWS has to offer. Services like Amazon EC2 Auto Scaling can help you optimize your workloads and save money on AWS.

By following these simple tips, you can get started on the path to optimizing your AWS savings.

Copywriting and editing by Jessica Zhou