Many Different Kinds Of Cloud, Very Big Piles Of Money

December 11, 2023 Timothy Prickett Morgan

From the very beginning, there have been many kinds of clouds. Yes, Amazon Web Services started out with the Simple Storage Service (S3) object storage service in March 2006, and quickly followed it up with the Elastic Compute Cloud (EC2) service, and even though it cannot decide if service is part of a product name or not, AWS set the stage for how infrastructure services (IaaS) and then platform (PaaS) and application software (SaaS) services would be layered on top of raw, virtualized iron.

Many observers thought that in the end, everything would go PaaS and SaaS and IaaS would just be a low-margin foundation for these services. Google, for instance, could not believe that anyone would want IaaS and its Compute Engine platform service had to be rejiggered to offer instances akin to those offered by AWS, Microsoft Azure, and other clouds. That was because there are still plenty of companies – millions of companies – that have unique software stacks that they cannot just throw away even as they go from on premises to cloud. And there are still others that think they can differentiate above the iron and below the application. And even SaaS and PaaS suppliers need to buy raw IaaS, too.

And so the bifurcation of the cloud market continues. As, in the longest of runs, at least some of your IT systems will, we think, reflect the distribution of the average at large. Which is why you need to know how cloud revenues by type are changing over time.

Let’s pick these numbers apart a little. First of all, growth across all of the cloud segments is forecast to be pretty good in 2023 and is also projected to be in the same ballpark in 2024.

This year, the analysts at Gartner are projecting that IaaS sales will rise by 19.6 percent to $143.9 billion, representing 25.5 percent of the market, and they project that IaaS sales will accelerate with 26.6 percent growth in 2024, reaching $182.2 billion and comprising 26.8 percent of the market.

PaaS revenues are in the same ballpark, rising 21.5 percent in 2023 to hit $145.3 billion and comprising 25.8 percent revenue share. Gartner is projecting for PaaS revenues to grow at the same 21.5 percent rate this coming year to $176.5 billion, and the PaaS share of the pie will grow by a tiny bit to 26 percent.

SaaS has been the biggest piece of the cloud market for many years, which makes sense because SaaS includes both infrastructure and platform services with applications running on top, and this year the market researcher is figuring that SaaS will bring in $205.2 billion (up only 17.7 percent) and will rise further in 2024 to just a hair under $244 billion (up 18.9 percent). With PaaS and IaaS growing faster, the share of SaaS is dropping ever so slowly, but still represents way more a third of the market and the largest slice of the cloud pie.

Business process as a service has 11.8 percent share of overall cloud revenues in 2023 and 10.7 percent share projected for next year, but it is growing more slowly than the PaaS, SaaS, and IaaS slices. Desktop as a service is a relatively tiny market, and while it is growing faster than the business process cloud slice, it is not growing as fast as the core cloud segments and therefore its share is dropping.

“Cloud has become essentially indispensable,” Sid Nag, a vice president and analyst at Gartner, said in a statement accompanying the figures. “However, that doesn’t mean cloud innovation can stop or even slow. The tables are turning for cloud providers as cloud models no longer drive business outcomes, but rather, business outcomes shape cloud models.”

This is certainly true. In the generative AI market, for instance, this has not only forced the adoption of GPUs from Nvidia (and soon AMD) to train large language models, but also to adopt InfiniBand for low-latency, high bandwidth networking and to adopt DPUs from Nvidia and others to offload network and I/O virtualization, security, and certain collective operation functions for the network. We think also that generative AI has driven the growth of IaaS services as well as PaaS and SaaS services on top of them because only the largest clouds have access to large numbers of GPUs to train large language models. As GPU supplies better meet demand, which they do not right now and probably will not next year, we think everything will normalize a bit and it is reasonable to expect for PaaS and IaaS to start outgrowing IaaS. Many companies cannot set up their own AI software stacks, and others simply don’t want to. They will adopt a pretrained model and retrain it for their own data and get going fast with generative AI.

This Issue Sponsored By

Table of Contents

Content archive

Recent Posts

Subscribe

Pages

Search