2029 will be a year to remember for AI: IDC

Mark your calendars now: 2029 will prove a great year for AI, according to Meredith Whalen, chief product, research, and delivery officer at market intelligence company IDC.

During her April 8 IDC Directions keynote, Whalen told the audience that the global IT industry is currently undergoing its largest growth cycle since 1996, with a current valuation of $4.2 trillion thanks to the rise of AI and agents.

“This isn’t just another cycle,” Whalen said. “This is a super cycle that we’re in and it is going to reshape the whole IT industry.”

Three years out. Whalen says 2029 will be the year that AI kicks into an even higher gear. That year, overall data center buildouts, for example, may have tripled from 2024 volumes, with an estimated $1.6 trillion spent on buildouts.

“It’s also when we start to see more organizations moving from that experimentation [phase] across into production, across multiple functional areas,” she said. IDC expects 2029 to be a “tipping point” for when capacity shifts from training models to delivering inference, or using trained AI models to make predictions or complete a task, at scale. Enterprise spend on infrastructure could reach $90 billion, up from $16 billion in 2023.

Bubble trouble. Not everyone is so certain about a glamorous future for AI investments. Schroders’ CIO Lens Q2, for instance, forecasts an AI bubble burst in the global equity market in Q3.

“Capex spending reverses, and consumer spending weakens as a mild recession causes unemployment to rise while tumbling stock markets hit sentiment,” Schroders Economics Group wrote in the quarterly report.

Here’s a tip. Building out AI capacity doesn’t come without challenges. Whalen said the role of SaaS vendors, for example, is evolving as applications no longer serve as the “primary interface for work” and more employees turn to AI. IT Brew previously reported on the ongoing “SaaSpocalypse” fears within the industry, which have been fueled by the belief AI will make SaaS applications largely obsolete.

“We know that there will be a few SaaS vendors that will be vying to be that agentic layer…but for the rest of SaaS players, you’re not going to be that,” Whalen said. “And that’s perfectly fine because your value is going to come in making sure that your application is part of the execution [layer] that’s required to get the work done.”

Device manufacturers, on the other hand, are dealing with constrained memory supply and rising memory prices because of the massive AI buildouts, Whalen said, adding that manufacturers should simplify their portfolios, and prioritize “high margin SKUs,” while selling value as opposed to features.