Tech analysts have forecast 9.8 percent year-over-year growth in worldwide IT spending in 2025, reaching $5.61 trillion and driven by continuing AI investments, despite “moonshot” projects seeing a high failure rate.
Our expectations for what generative AI can and will do are starting to come down…
Global research firm Gartner stated that overall IT spending growth is expected to accelerate from 7.7 percent in 2024, driven by steady economic expansion and growing interest in AI.
Spending in datacenter systems is projected to rise by 23.2 percent in 2025, reaching nearly $406 billion, according to Gartner. While this marks a slowdown from the 39.4 percent growth recorded in 2024, it was still a strong number, John-David Lovelock, distinguished VP analyst at Gartner, told The Register.
“Servers grew 68 percent in 2023. You can’t keep that kind of pace up. We are already on track for servers to triple in spending from 2022 where we spent about $130 billion, to 2028 where we’re going to spend $447 billion, which is astronomical,” he said.
Spending on AI-optimized servers in 2025 is projected to be twice that of traditional servers, reaching $202 billion.
However, despite the experimental deployment of AI in business during 2024, businesses remain uncertain about the true benefits of the much-hyped technology, according to Lovelock.
“The GenAI work that was done in 2024 — the transformational stuff, the moonshot projects — those are pretty much going to be over for 2025. They were maybe a little under-invested in and there were pretty high failure rates,” he said.
As a result, CIOs were beginning to think that the assumption that AI would create a transformation in business might not work out, Lovelock said. “Our expectations for what generative AI can and will do are starting to come down.”
Famous for its Hype Cycle, which tracks how emerging technologies progress through the peak of inflated expectations, the trough of disillusionment, and ultimately to the plateau of productivity, Gartner now sees GenAI on a downward slope.
“We won’t make it to the trough until 2026, but 2025 is going to be a year of the slide,” Lovelock said.
User organizations have struggled to see value from GenAI because they have been too focused on starting with a model and hope to develop use cases from it, he pointed out.
“When they are working with existing systems and layering in GenAI functionality, then they did slightly better. But, as [OpenAI CEO] Sam Altman said when he launched GPT-4, they still aren’t all that good. We had a lot of failures,” Lovelock said.
“A lot of it was down to the data the organizations had or they would think ‘Is this all we get?’ even when they could get to 90 percent of it. Suddenly, the cost and value started not to align, and then they would realize they don’t have the people to maintain and run this after it’s in production, even if they could get it to production.” ®
