The growth trajectory of hyperscalers (AWS, Azure, GCP, etc.) has been incredible, as demonstrated by the projected 28% CAGR for the period between 2022-2030, but there are signs of a slowdown.
Recently, AWS has seen their growth cool to 28% down from 39% in the same period the year prior. The story was the same for both Azure and GCP. The core question is what is leading to this slowdown? There is evidence that companies are starting to re-think their cloud strategies.
37 Signals – An Impressive Cost Savings Story
37 Signals is a midsized SaaS company that operates both Basecamp (a project management application) and Hey (a reimagined email client). They made an announcement in October of 2022 that they were migrating off both AWS and GCP.
They had been hosting their platform for over a decade on hyperscalers and have run under a variety of arrangements during that time from running on bare VMs to Kubernetes. The conclusion they reached after 10 years was that “Renting computers is (mostly) a bad deal for medium-sized companies like ours with stable growth. The savings promised in reduced complexity never materialized. So we're making our plans to leave.”
This January, they announced they spent $3.2M on cloud services in 2022. This was after a massive amount of work was done to optimize their spend “The ops team runs a vigilant cost-inspection program, with monthly reporting and tracking, and we’ve entered into long-term agreements on Reserved Instances and committed usage, as part of a Private Pricing Agreement. This is a highly-optimized budget.”
So what does the future hold for 37 signals? All in, they expect to pay $840K a year go forward through purchasing their own equipment and collocating it. This represents a 64% reduction in spend. In addition, 37 signals expects other benefits such as “much faster hardware, many more cores, incredibly cheaper NVMe storage, and room to expand at a very low cost”
I Thought Hyperscalers Were The Most Cost Effective Approach. What Changed?
Fees, fees, and more fees. The Wasabi Cloud Storage Index reveals that customers using cloud storage services from hyperscalers can expect 48% of their total bill going towards fees. Fees include: API requests, egress, data retrieval, replication, transfer acceleration, object tagging/minimum object size, minimum retention.
We are also seeing increased margins. In Q1 of 2022, AWS’ operating margin is now 35.3%, compared to 29.8% for the fourth quarter of 2021 and 31% a year ago. This has slowed in 2023 with a reported margin of 24% (the lowest since 2017) but the question remains will AWS settle for lower margins or will they continue to raise fees to claw back the margin degradation?
We believe AWS will not consider these lower margins the new normal as AWS has fueled much of their overall profitability. In 2022, 74% of Amazon’s operating profit came from AWS and expect AWS to make up for margin loss through increased pricing or continuing to eek out more usable life from equipment (AWS, in 2022, added one year of useable life to both their servers and networking gear)
The limited menu offered by hyperscalers can also drive-up pricing. Hyperscalers couple memory to vCPUs so to increase your vCPUs you are also going to get more memory whether you need it or not. This means you are often paying for more resources than you need for a use case and are forced into a larger instance type to meet your resource requirements for a particular “component” (vCPUs or memory).
Are Hyperscalers The “Right” Choice Ever?
There are 3 primary use cases for hyperscalers:
1. Startup mode – when you don’t have the capital to purchase your own equipment with an uncertain customer base and revenue streams hyperscalers are a great place to start. You can set up a very lightweight environment that can quickly scale if you catch lightening in a bottle (which is a good problem to have).
2. Peaky workloads – does your workload have wild swings or towering peaks in usage? Then hyperscalers are a great fit. If your baseline is a small percentage of your peak needs, it makes little financial sense to scale out for these infrequent peaks in usage instead you can use autoscaling features in the cloud to provide a performant experience to your customers during peak usage and then scale down quickly when your baseline usage returns.
3. Cloud-Native Applications (Serverless) – do you plan on releasing applications that are built using cloud-native development architectures such as Azure Functions/Logic Apps, AWS Lambda/Step Functions, etc.? In this case, you have no other choice but hyperscalers. Prior to going this route though you should ask yourself what is your motivation for using this development approach?
Vendor lock-in becomes even more problematic with this approach and you should have a clear understanding of the benefits you expect to achieve going this route.
What Are The Alternatives?
The good news is that we are seeing more viable alternatives than we have seen before in the space. You can divide your potential approaches into 4 categories. The chart below represents who is responsible for what in each arrangement.
If you already have moved your workloads to a hyperscaler and don’t want to get back into the “hardware” maintenance game, Infrastructure as a Service (IaaS) is a great fit.
Going from on-premises to the cloud? Consider both colocation and IaaS as an option along with hyperscalers.
Many co-location and IaaS providers offer professional services that you can take advantage of to create a “blended approach”. Want to collocate but don’t want to be responsible for bare metal servers but want to continue maintaining your storage? The good news is you can create a combination that works for you by blending your own workforce and professional services.
One Final Thought
We will leave you with one final thought from David Heinemeier Hansson, the co-founder and CTO, of 37 Signals…."Any…business…with stable workloads that does not benchmark their rental bill for servers in the cloud against buying their own boxes is committing financial malpractice at this point."
Want to ensure you are not committing financial malpractice? We can help. Reach out to us and we can help benchmark your on-premises or hyperscaler spend to a variety of colocation and IaaS providers to ensure you are maximizing your infrastructure spend.