Hardware-as-a-Service Isn’t the Future of On-Prem IT

Every hardware vendor in the world now is pushing for Hardware-as-a-Service (HaaS). Even though I agree that software-as-a-service (SaaS)-like subscriptions are interesting consumption models for hardware, I don’t think it will make any difference for the general outcome of enterprise IT and, more in general, it will remain one of the many options… not even the predominant one.


Hardware-as-a-Service is one of many ways to describe a new subscription-based model to purchase hardware, software, and services. It tries to mimic cloud flexibility and to move more investments from CAPEX (capital expenditure) to OPEX (operational expenditure). All vendors now have similar models and the differentiation is minimal.

When you talk with vendors, they all love this sales model and talk about increasing business made through it. But let’s be honest, when you look at the numbers, for the most part they are converting existing business and not really increasing revenues. Some of them talk about “double-digit” growth in 2019. From my point of view, they are overly excited. Just to make an example, last week I heard a CEO of a primary hardware vendor talking about double-digit growth for sales made through subscriptions, something that last year was accounting for 0 or just a few percentage points of the total revenues. I’m not a finance guy, but for me, even +79% of almost nothing is still almost nothing.

It Is a Great Model (In Theory)

But let’s be positive here. In theory, Hardware-as-a-Service is a great deal for the user. Private data center won’t disappear anytime soon, and trying to switch from traditional purchasing models to something more flexible isn’t a bad idea.

Pay-as-you-go is what everybody wants. No overprovisioning, no commitment, quick resource allocation and deallocation, flexibility. Who doesn’t want it? But unfortunately this is not exactly true, and financial wizardry can’t give you the cloud. Cloud and on-prem are totally different beasts and trying to make them resemble each other doesn’t bring any benefit to the user.

A Quick Comparison (In Practice)

The big advantage of the public cloud is extreme flexibility. You pay for everything by the minute, with unlimited resources at your disposal distributed across the world, with all the services you may think of, and ready to use. You pay for it, for all of it actually. Cloud is expensive.

The main reason you go to the cloud is that you want to be agile no matter the cost. Fast results, freedom to fail and restart quickly, to avoid any sort of complex and risky capacity planning, no upfront investment, and so on, are what draw users to the public cloud. Can you see any of this in your on-prem data center? Can Hardware-as-a-Service change it? The quick answer is simply no, you can’t.

Your on-prem data center is rigid and paying a lot more for hardware to get a very limited improvement in flexibility doesn’t make any sense. Here is why:

  • Minimum order commitment: As far as I know, vendors have yet to apply this model for small installations or customers. In most cases, it is not clear what the table stakes are, but to start playing you have to put some real money on the table. From an informal survey I conducted last week, it is hard to get a quote for hardware that doesn’t have a counter value of $100K or more. Meaning that small projects, SMEs, and departmental IT are out of the game in many cases.
  • Minimum time commitment: Another issue comes from the minimum time commitment for this kind of subscription, which is one year in most of the cases. Meaning that you need to stick with what you purchased for at least one year, even if you made a mistake or your project fails. You can expand it of course, but you can’t change what will become the core of your infrastructure. BTW, you can’t shrink it as well in the first year.
  • Delivery time: we talk hardware here! Even in the most positive scenarios, getting new equipment delivered to your data center will take days, but more likely weeks. It’s not exactly as quick as in the cloud, is it?
  • Datacenter time: Did you change something in the way you manage your data center lately? Because the same process inefficiencies you had in the past will affect the installation and configuration of new hardware as always. Including power and network cabling, and anything in between that and the moment you can switch on the new hardware.
  • Decommissioning: Nobody talks about it, but what happens when you want to give back entire parts of the hardware you bought as a service? Good luck with that.

As I wrote above, all the alleged flexibility described above has a price… because, at the end of the day, it is clear that someone has to pay for it and, usually, it is not the vendor.

Takeaways (and Common Sense)

Subscription-based models make sense on several occasions, it is just that I doubt they will become the predominant way to purchase hardware. In some cases, it could be better than traditional leasing, renting or other financial wizardries, but I’m sure that this wasn’t the goal. Also, there are some niche use cases, such as the case of MSPs, where this kind of sales model makes more sense in my opinion (but this could be a story for another day).

All this Hardware-as-a-Service craziness will fade over time. The beauty of the hybrid cloud is the ability to move applications and data where they fit at best, depending on business, financial, and technical needs. Why should we aim for a more expensive on-prem IT without actually getting a level of flexibility that is nowhere close to the cloud? Isn’t it better to have rigid IT at a low cost and flexible IT at higher prices? Isn’t it better to have multiple tiers for our IT services?

Containers, Kubernetes, SD-networking and new storage technologies are re-defining the concepts of application and data mobility. Data and applications are now able to move freely and (almost) seamlessly. Users can start in the cloud and take advantage of flexibility, agility, and infinite scalability, then they can move applications and data to a less expensive on-prem IT infrastructures when it presents a stable behavior for which capacity planning is easier to do and cost becomes a key aspect.