Cloud Revenues Will Bust Through $1 Trillion Before Too Long

Cloud is an attitude as much as it is a consumption model. So is parfait, and as we all know from Shrek, “everybody loves parfait.”

We love to track the parfait of cloud revenues, from the lower chocolate layer of server, storage, and networking levels commonly abbreviated IaaS, to the middle butterscotch layer of platforms and frameworks shortened as PaaS, to the French vanilla layer of business process automation called BPaaS, and finally to the homemade whipped cream or ice cream of application software known as SaaS.

The IT market has been trying to create a parfait of hardware, software, and services spending since the dawn of the commercial computing era six decades ago, and we have certainly perfected it in recent years. It was only because companies were control freaks – for both their infrastructure and their balance sheets – and were looking for the cheapest way to run very expensive infrastructure that we had many decades of on-premises, capitally expensed corporate computing. IBM (and other companies) charged so much for renting their gear that companies decided to build glass houses and show off their economic might by owning their own mainframes rather than renting them, and Big Blue was compelled by two antitrust lawsuits with the US Department of Justice to let them do that at a reasonable price. And thus, on premises computing was born.

With the commercialization of Internet technologies in the mid-1990s, and massive investments in fiber optic networks that spanned the globe, it became technically and economically possible to think about geographically distributed computing and this thing we used to call “application service providers,” which was an early attempt at delivering SaaS. The infrastructure wasn’t cheap enough or good enough, and this largely failed. There were some exceptions and some ASPs made it – Salesforce being perhaps the biggest exception.

And with the advent of Amazon Web Services in early 2006, hyperscale was married to utility-style infrastructure that could be turned off as easily as it could be turned on and the cloud as we now know it, with all of its layers, was born. To be fair, Google Compute Engine started out as a PaaS and the search engine giant and cloud wannabe had to be dragged kicking and screaming into exposing its infrastructure at a more primitive IaaS level that customers had been taught by AWS and then later Microsoft Azure to expect.

And so, we have this cloud parfait. Some eat all the layers, some just go for one or two layers. But the import thing is you can eat whatever layers you want, and virtually as much or as little as you want, and you only pay for what you eat.

Not everything can move to the facilities created by cloud builders large and small, of course. And while it may feel like everything will go to a cloud, as far as we can see, there will be plenty of on-premises IT gear sold in the old-fashioned way and put on the capital expenses part of the balance sheet instead of being booked as operational expenses as cloud capacity is.

Gartner has been tracking sales of IaaS, BPaaS, PaaS, and SaaS for more than a decade, and even tracks this puny little category called Desktop as a Service, or DaaS, that is still a multi-billion dollar annual business but which has not taken over the world despite many generations of “network computers” – remember those? The company has just released its estimations of cloud revenues by type for 2023 and is providing an updated forecast for 2024 and also pushed it out to 2025. The forecast actually goes out to 2028, but that data is not public. What is clear is that if current trends persist – and there is absolutely no reason to believe they will not – cloud revenues worldwide will surpass a $1 trillion annual run rate somewhere around June 2026.

This will be a remarkable feat.

What is interesting about this cloud spending parfait right now and into the immediate future is how relatively evenly distributed the revenues are for the IaaS, PaaS, and SaaS layers. Over time, if more and more customers move software from their own datacenters to the clouds, the SaaS layer, which is largest, should grow faster. But right now, thanks in part to the generative AI boom, the IaaS and PaaS layers of the cloud builders are growing faster than SaaS. It seems unlikely that either IaaS and PaaS will be able to catch up to SaaS in terms of absolute annual revenue levels, and Gartner’s forecasts certainly show SaaS outgrowing IaaS and PaaS from 2023 through 2025.

This historical chart, for which we compiled the Gartner data going all the way back to 2015 (with the adjustments to annual figures as each year progressed, which is a bit of a pain in the neck), gives you a sense of how the different levels of the cloud parfait are changing over time:

The fun bit in there is that BPaaS was a bigger part of the market a decade ago than SaaS, and IaaS was half the size of SaaS and PaaS was a relatively tiny slice. A decade ago, Gartner added in a category called Cloud Management and Security Services to its overall cloud revenues, which it removed two years ago and which added another 6 percent or 7 percent to the market. Gartner also used to add in cloud advertising to the cloud revenue mix, and took that out in 2018 because, quite frankly, it was silly to have that in as infrastructure.

Anyway, including the 2024 and 2025 forecasts, PaaS has grown from 4.2 percent of all cloud revenues in 2015 to hold around a steady 25.5 percent from 2023 through 2025. SaaS was around 35 percent of the market in the mid-2010s, grew to 40 percent in the late 2010s and early 2020s, and now is back down to 35 percent in the forecast for 2024 and 2025. IaaS was only 17.9 percent of cloud revenues in 2015, went up to the 20 percent range for a few years, dropped down a few points for a few years, and then moved on through 25 percent and will start pushing up to 30 percent of revenues in 2026 or so if current trends persist.

This cloud spending can be thought of as a kind of rough breakdown of an IT budget in the datacenter. Compare your own spending at for these layers of the IT stack to those of the cloud consumers. Do you spend more of your budget on raw infrastructure, platforms, or application software?

Sign up to our Newsletter

Featuring highlights, analysis, and stories from the week directly from us to your inbox with nothing in between.
Subscribe now


  1. A bust is a woman’s breasts or a head and shoulders sculpture. It is also poor slang for breaking either physically or financially. The proper word is “burst”. I did not read the article because of the low educational level this implies and so as not to waste my time.

    • You bust through a closed door. Low educational level, indeed. I am tempted to tell you the many ways you can go verb yourself. But you are not worth the effort.

Leave a Reply

Your email address will not be published.


This site uses Akismet to reduce spam. Learn how your comment data is processed.