How Much Software Budget Does AI Drive?
December 15, 2021 Timothy Prickett Morgan
Fake brain-like software based on neural networks is all the rage now, and probably will be for the next decade or so, but you may be surprised at how much money is being spent on software that has some sort of AI embedded in it and also to learn what a small piece of the overall IT market that this AI software represents at this point in IT history.
The market researchers and prognosticators at Gartner recently tried to case the AI software market, and looked at how much money the top five use cases for software that have AI embedded in them are generating and then added all of the many other use cases together. Here is what sales look like in 2021 and how they are expected to grow in 2022:
This is obviously very hard to track, and for some very good reasons. First of all, these AI spending numbers are relatively low because so much of the production-grade AI software, particularly relating to machine learning training and inference frameworks and libraries, is either open source or freely distributed along with hardware such as GPU accelerators. (The latter is true of Nvidia, the dominant supplier of AI platforms on Earth right now as well as a maker of GPUs for compute and graphics.) The second reason these numbers are not larger is that a whole slew of spending on AI software is done by the hyperscalers and cloud builders, who in turn sell services based on their own AI platforms. None of this is counted strictly as an AI software sale, of course. And perhaps equally importantly, we have no idea of the amount of AI software investment that hyperscalers, cloud builders, large enterprises, and supercomputing centers are making in AI software for all kinds of use cases.
According to a recent survey by Gartner, 48 percent of chief information officers say that they have either deployed AI technologies – including machine learning applications driven by neural networks but not limited exclusively to this approach – or will do so within the next twelve months.
Gartner adds that even if AI software is fairly large and growing reasonably fast, the deployment of AI software is much more limited, and most organizations are still only at the experimentation phase with AI. So if you think you are behind the eight ball on this, you really are not. Gartner, in fact, expects for it to take until the end of 2025 before half of the organizations in the world to reach what it calls the “stabilization stage” of AI maturity. As the AI use cases become clear and the AI systems find new use cases – as has happened with ERP enhanced with customer relationship management or supply chain management software two decades ago when these ideas were new – the AI software market will grow. At some point, AI algorithms will just be the way that a lot of software is done, just like these days half of the IT infrastructure sold worldwide is really for private or public clouds.
“Successful AI business outcomes will depend on the careful selection of use cases,” explained Alys Woodward, senior research director at Gartner, in a statement accompanying these figures. “Use cases that deliver significant business value, yet can be scaled to reduce risk, are critical to demonstrate the impact of AI investment to business stakeholders.”
While the $51.5 billion in AI software sales in 2021 and the $62.5 billion in sales anticipated in 2022, which represents a 21.3 percent growth rate, is a lot of money compare to, say, the size of IBM’s entire Power Systems business, it is a pretty small compared to global IT spending figures put together by Gartner, which we covered last month. Gartner reckons that global IT spending will be $4,241.6 billion in 2021, and will rise by 5.5 percent to $4,474.2 billion in 2022. If you do the math, AI software spending was officially 1.21 percent of overall IT spending in 2020 and will rise to become 1.4 percent of IT spending next year – at least int eh way that Gartner carves up the IT market. So much of what we do in IT has nothing to do with AI, at least for now. It will be interesting to see how this looks a decade from now – if we are all not living in the Metaverse or Omniverse with wires up our nodes by then. Count me out on that one. I like real reality.