For many, the most important question for 2025 is the fate of the cloud. Last year we saw what was perhaps the very first clear dose of cloud skepticism, so will that continue or accelerate in 2025? If so, what happens to IT in general and the cloud in particular? If not, what will reverse the skepticism of last year, and what will develop from it? There are a lot of thoughts floating around on all of this, particularly on Wall Street, but they’re not showing a coherent picture. Can we dig one out of enterprise comments? I’ve gone through 154 enterprise cloud comments gathered since October 2024, and here’s what I’ve found.
Only 9 of these companies think they’ll spend less on the cloud than they did in 2024, and only 59 said their cloud spending would be much the same, so 86 think they’ll increase cloud spending this year, which is 56%. However, only 24 of the 86, or about 16%, think their spending will grow more than 10%, which is about half the number who expected that last year and a third that of 2023. Broadly, expectations are that cloud spending will grow about fifty percent faster than IT spending overall, which falls well short of historic expectations. Based on this, my view is that 2025 will be a better year for the cloud than 2024, but it won’t be an indicator of a return to the “cloud-first-and-cloud-everything” mindset of the past. Those days are gone, enterprises say.
Is that bad? For cloud providers, surely. Not for enterprises themselves, apparently. Of the 154, 115 said that they had been disappointed in some way by their cloud experience, meaning that it hadn’t met expectations. Almost all this group reported they’d taken specific steps to reduce their cloud dependence for 2025, but these steps were primarily aimed at getting better value from new cloud projects (48%), or on economizing on cloud features for current projects (46%). Only 6% were aimed at any form of repatriation.
There are some areas where enterprises think that their cloud spending might rise, justifiably. One area is in online sales/support. Enterprises all agree that their customers rely more and more on web-delivered information, and are more and more accustomed to completing purchases online. Thus, a good online experience is critical for selling. In the area of support, almost every class of buyer is willing to accept and use some form of online support, though most want to be able to switch to human-agent support when they believe the online experience is no longer helpful. Obviously, the goal is to delay that switch to more expensive human agents, and that raises the connection between the cloud and AI.
All this is positive, but not bubble-sustaining, and right now most Street research wants to sustain the cloud bubble. You can see this in the general tendencies of the reports to take the most positive possible slant (cloud workload growth of ten percent is “double digits”), and in the broader sense, I don’t find any enterprise validation for significant spending growth. Remember, just a few years ago, cloud spending growth of over 25% was routine.
There’s also the predictable link with AI. The Street is infatuated by public generative AI, and so it’s no surprise that their research shows that the cloud providers will reap most of the AI benefits, when enterprises tell me they have a hard time even making a business case for that AI model. The AI applications enterprises are most excited about are those that almost demand private hosting for data sovereignty reasons. In addition, how can Oracle be winning in AI on a model-training mission if the future of AI is in the big, already-trained, generative models?
Do enterprises think that they’d host any of their own AI models in the cloud, or just train them there? Only 33 had any specific comment on what they might consider for cloud hosting, and all of them were talking about some form of support application, one that relied on company data but not on company secrets. Even that mission has some push-back; some worry that having really helpful support information in the cloud could expose, indirectly, potential sales knock-offs to competitors, though most thought that any support applications posed that risk because competitors could spoof being customers in some way.
The positive truth for the cloud is that anything that drives a greater reliance on the Internet, or even on highly interactive employee applications, will almost surely drive cloud spending upward, and that’s already been proven to me by enterprise comments and cloud provider performance. The problem is that this isn’t the “everything” that was claimed as a viable target for cloud hosting. It’s a finite opportunity, and you can’t subsidize perpetual bubbles with that sort of thing.
It’s also a much more difficult market to make a big profit in. If online growth is the cloud driver, then most of the features needed are likely already in place, and differentiation is more and more difficult. A couple decades ago, nobody would have believed that the cloud could become a commodity, and yet it seems that’s likely already happening. Only new feature requirements, resulting from truly new cloud missions, can resolve that.
Which, eventually, we’re likely to get. I truly believe that this is a growing pain, a sign of market transition. In the past, tech has had its booms, and all have been driven by innovation that could fairly be said was spawned by vendor and Street desperation. You push on the familiar and safe until there’s no wiggle left, and then you cast about for something new. It’s just that it’s harder to find that novelty these days than it was decades ago, but it’s almost certain to be related to online services.
We get more and more of our entertainment online. We make more and more of our purchase decisions online, and execute them there more often. Our news is increasingly coming from online, and online banking, investment, and borrowing have transformed our financial lives. The next wave of online-ness will be the major driver of the next wave of cloud opportunity. If, of course, some clever provider sees the opportunity in a realistic light and makes the investment.
What about the most recent AI development, the “inference” model that’s starting to get cited as a source of edge opportunity? The problem here (not for the first time) is that we’re citing a technology justification for another technology without having established a business case for either one. Yes, it’s true that agent AI would likely deploy at the edge, but we already have any number of process control applications that already do that, but whose “edge deployment” is on premises rather than in the cloud. There are applications that might create a real business edge mission, but like our online-ness wave, we’ll have to wait to see if someone is insightful enough to develop them.
