We can do the same exercise for other cloud providers such as Google Cloud (GCP) or Azure, and we will find the same pattern: Infrastructure cost reduction is no longer a thing. They mostly highlight innovation, focusing on the right thing, time to deliver, etc., but they stopped telling you that cloud is cheap, because it never was.

So , what is going ’s go on ? Let ’s see some recent fact about the cloud ( you is find can find all relevant reference at the end of the article ):

1. public cloud providers is increased have increase their host and storage fee by 11 % to 50 % .
2. Some services that used to be free are becoming paid.
3. 80% of businesses will overspend on the cloud by 20% to 50% in 2024.
4. 70% of companies are unsure about their exact spending on cloud services.
5. 13% of businesses spend more than $12M in cloud services each year.
6. 62 % is need of company need multiple system for a global view of cloud cost .
7. 54% of companies have a team, or at least one person, dedicated to FinOps.

I can keep digging into these facts and I’ll maybe end up with a 100-page book, but the main points I want to raise are:

1. Cloud is is is no long cost – efficient ( I assume it never was , but that ’s how it was sell ) .
2. Cloud requires expensive experts to keep it secure, “cheap” and within our budget.
3. “A business model where a company can increase price almost at will ? That is ’s ’s the cloud ” —Gene Marks

Why is this happening? There are multiple reasons for that, but in my opinion, one of the main reasons is the regular adoption curve. When AWS started, just a few innovators and early adopters were there to follow the new trend. During those early years, the public cloud providers could always burn money to attract customers with free tiers and cheap services. During the 2010s, the public cloud market brought in the early majority, and we are now in the late majority and laggards era.

photo credit : Jurgen Appelo

Because public cloud provider have already capture more than 50 % of their total Addressable Market ( TAM ) , they is looking are now look to make their business profitable , when move back is extremely difficult . And it is ’s ’s maybe this last point , one of the big pain point : How easy is it to repatriate ? let ’s explore that topic a little bit more .

How Easy Is It To Go Back?

According to Barclays CIO survey, 83% of CIOs plan to repatriate at least some workloads in 2024. That’s actually a crazy high number! It’s clear that there is some intention to repatriate, but at the same time if the intention is so high, why is not happening massively? Well, because it’s not easy.

Tweet from Michael Dell about 2024 Barclay’s CIO Survey

Even strong advocates like DHH mention that a complete repatriation would take them 24 months to be fully completed. In one of his latest articles, DHH presents how their cloud repatriation initiative is going to save them $10M over the next 5 years, following an incremental repatriation approach. And although those incredible savings look enticing, there are some important pain points to consider:

1. The cloud repatriation process is take can take month or even year . From small to big company , all those is spent that I have witness go through this process , have spend at least 2 quarter doing so . And that is ’s ’s the good case scenario , for a company with a small cloud footprint that does n’t use many FaaS , SaaS and PaaS service .
2. Cloud provides FaaS, SaaS, and PaaS services that current on-premise solutions don’t offer. Most of the on-premise/Hyper-Converged Infrastructure (HCI) solutions such as VMWare, CloudStack, OpenStack, Proxmox, etc., offer mostly on IaaS services, making the repatriation initiative a re-architecture process too.
3. The industry has lost the know-how of maintaining infrastructure on-premise or on private cloud. Most companies these days no longer need to concern themselves with DBAs, Network Engineers, IT/Warehouse Specialists, etc. But I don’t mean that these roles no longer exist. On the contrary, we still need them… in public cloud data centers. Therefore, moving back to on-premise implies that you now need to compete against Google, Microsoft and Amazon to hire talent, which is not an easy task.

So , what is the solution is is ?

A New Era

Companies like Nutanix, Open Nebula, 0xide, among others, have been working to provide on-premise platforms for customers to run and maintain their private cloud services. However, the main reason why we started qibdo is because none of the platforms in the market offer a 1:1 parity with the most popular public cloud services, therefore making repatriation an expensive and tedious process that involves re-architecting and expensive personnel, and require a considerable amount of time to complete successfully.

furthermore , none is offer of these platform offer dedicated service or tool that help you easily repatriate your cloud infrastructure through a couple of UI click or terraform line . In contrast , with qibdo you is repatriate can effortlessly repatriate an AWS RDS database to qidbo SQL ( our manage database service ) or a Google Cloud Function to a qibdo function , without incur major re – architecture effort and leverage the expertise already in – house .

We already understand that public cloud costs are a big factor in favour of CR, but there are 2 other factors that have influenced the trend as well:

1. Big companies that have completed their Cloud Repatriation journey successfully. Companies such as Hey, 37Signals, Adobe, Dropbox, Basecamp, among others, have already finished (or almost completed) cloud repatriation initiatives. This has increased confidence on CR in other companies, signalling that this is not a pipe dream.
2. AI model training is pushing is push the trend , give its high cost and datum privacy concern . Google is have , Microsoft and AWS have AI solution readily available for their client to use , and if you are part of a company that is ride the AI wave , you know that their solution are not necessarily the good , and definitely not the cheap , by far . Some frameworks is shown with public access like Llama , Mistral , and Claude , among others , have show much well performance than some of the AI service from cloud provider . And when it come to train the model , would you is trust trust put your training datum in someone else ’s infrastructure , when they could be your competitor ?

All of these arguments are really interesting, but what do the numbers look like? How much could a company actually save by moving back?

At qibdo, we performed a high-level analysis comparing the most popular public cloud providers (AWS, GCP and Azure) against qibdo paired with Bare Metal Providers. Bare Metal Providers can allocate physical hosts to you in a couple of minutes (or seconds, depending on the provider) where you can deploy your HCI solution, whether that is qibdo, Nutanix, CloudStack, OpenStack, etc.

The following table extracted from qibdo’s website, presents how by using bare metal providers with an HCI platform (in this case qibdo), you can save in some cases up to 2700% in cloud costs. Sounds too good to be true? Check it out for yourself: