There is currently no shortage of articles talking about how more and more companies are moving back to on prem, citing that the cloud has failed to deliver the cost savings as promised. I've even had several conversations with close friends about their personal experiences with the cloud just not living up to their expectations. I believe that these arguments are failing to look at the whole picture, and compare all of the services provided.
Cloud providers offer a wide array of features and services that often go unnoticed but significantly contribute to their value. Their platform itself is arguably the biggest underappreciated service. A cloud platform is a tremendously useful tool, allowing you to leverage orchestration automation, inventory management, configuration management, and streamlined deployment pipelines. All of this from a centralized, API-driven environment. You also get to utilize their state of the art, highly efficient network infrastructure, and DDoS protection. Beyond that, you're tapping into a global-scale infrastructure: cutting-edge networking, DDoS mitigation, hardened data centers with robust physical security, and highly redundant power and cooling systems. You get access to enterprise-grade hardware, optimized storage, and high-performance compute. You benefit from their world-class software stack, their databases, their ML/AI platforms, and the full ecosystem of tools they've built to simplify complex operations.
But perhaps most importantly, you gain access to the people: security teams, compliance experts, SREs, and support engineers. All these people working around the clock to monitor, maintain, and improve the platform so you don't have to. Cloud providers attract some of the most talented engineers in the world, and I shouldn't say some because they employ and army of them...so many engineers. You could absolutely replicate all of the offerings provided by a cloud yourself but at what cost? Simply running 1000 vms on prem might be cheaper than running them in the cloud, but what about the half dozen dev teams you need to hire to build your orchestration platform, and the network architect and operations teams you need to build your network? Don't forget reliability. You're going to need an SRE team...you're going to need one hell of an SRE team to come even close to the same level of reliability as one of the major cloud providers. Five 9s is not cheap.
This leads to another issue I have with the rhetoric around the cloud being “too expensive.” In many cases, companies aren't going all-in, they're just dipping a toe. They migrate part of their infrastructure to the cloud while keeping critical systems on-prem. As a result, they end up needing cloud specialists in addition to their existing operations staff. Instead of streamlining, they increase their operational overhead, managing two parallel infrastructures. This results in doubling up on tooling, processes, and expertise.
The reasons companies give for resisting full cloud adoption are often ludicrous. Sunk-cost, security, mistrust, risk averse. Probably the most common is fear of a "single point of failure." But the cloud is no more a single point of failure than your own internal IT ops team. The idea that a mid-sized company's in-house team can deliver higher availability than dedicated cloud provider is, frankly, laughable.
Sure, in theory, any company could build an ultra-resilient infrastructure. However, if they were actually paying the full cost to do that correctly: redundant data centers, 24/7 on-call teams, global failover, continuous testing; then they probably wouldn't be complaining about cloud prices.
Should you put all your eggs into one basket? Maybe not, but for some people you should, because your basket probably sucks and chances are you can't even carry two baskets. What's better: carrying your eggs in a basket with a 99.999% chance of success, or paying 2x the cost of the eggs for a second basket and then having a higher chance of dropping both?