While generative AI will drive up spending on enterprise software worldwide, most of that revenue will be captured as a de facto “tax” on developers, according to Gartner distinguished VP analyst John-David Lovelock.
Gartner recently forecast that IT spending will rise 7.5% globally in 2024 year over year, reaching around $5.26 trillion. While it projected the data center systems segment will grow by over 24% this year alone in large part due to demand for generative AI processing power, software wasn’t far behind. That category, which includes AI inferencing, is projected to grow 12.6% to over $1 trillion.
In an interview with IT Brew, Lovelock explained why software developers don’t necessarily stand to reap benefits from that surge in software spending.
This interview has been lightly edited for length and clarity.
What do you mean by generative AI being a “tax”?
So if we take a fictional company as an example—say, TomSoft. And you play in the enterprise resource planning, supply chain management, customer relationship management space. You have until the end of 2025 to embed GenAI into your product… Unless TomSoft was smart enough to already have been developing a GenAI model, you have no opportunity to develop one by that deadline.
Most companies missed that window. Most companies are going out and partnering with some type of GenAI model provider—whether it’s one, two, or going to a consortium or a collaborator, or an orchestrator or conductor service, they don’t have their own GenAI. But they are betting a third-party version of it to their clients, and then they’re able to say, “Here’s TomSoft plus GenAI.”
Now you can charge a premium for this thing, have tiered pricing, you could do the CoPilot thing and say it’s an extra $25 a month. Or maybe what you’re doing is charging tokens and saying the software itself is free, but just pay the tokens for using it. Either way, you’re in front of your client saying “Here’s generative AI functionality that I didn’t create, but I’ve embedded”...
Top insights for IT pros
From cybersecurity and big data to cloud computing, IT Brew covers the latest trends shaping business tech in our 4x weekly newsletter, virtual events with industry experts, and digital guides.
From that point of view, GenAI to you is a tax. You have to have GenAI, like a tax.
Do you see this driving an increasing share of software spend to a smaller handful of developers—the ones that have already been dominating this trend?
Their revenue is certainly going to go up, but their costs are also going to go up. They do not have control over their cost of goods sold as long as they’re using a third-party generative AI model or models.
Who are the primary beneficiaries?
There is the specialized Infrastructure-as-a-Service market—ones that run on GenAI optimized servers. That’s a brand-new market up to this point. We have had CPU based systems, and Kubernetes, and arrangements of bare iron, and API calling, and VMware instances. But it has all been sort of on these cheap, bespoke [original design manufacturer] servers that have been built for the hyperscalers.
GenAI is a new set of servers. It has a very new price point—a million dollars a rack is not uncommon. It has a different functionality, and it brings a different revenue…It’s going to come to the traditional three [Azure, Amazon Web Services, and Google Cloud] in the Infrastructure-as-a-Service base. They’re the only ones with the kind of money to build out these massive new data centers, or massive new server rent.
On the other side are the companies that actually provide the models, build them and then inference them. To Gartner, this is a brand new IT market.