Data center capex could top $1 trillion by 2027 thanks to AI

If there is one word to describe data center capex spending in 2023 it’s probably “meh.” It seems as though that’s about to change in a big way, thanks primarily to investments in artificial intelligence (AI) infrastructure.

As always, analyst estimates of where things landed in 2023 vary and there’s still a bit of uncertainty about 2023’s figures because Q4 reports have yet to be made. But Dell’Oro Group estimated global data center capex grew just 4% in 2023, rising from around $241 billion in 2022 to roughly $251 billion. That compares to 15% growth from 2021 to 2022. 

New Street Research (NSR) put the 2023 capex number a bit higher, at around $300 billion, with $125 billion coming from the top four hyperscalers. AI-related capex specifically was pegged at roughly $90 billion, by NSR’s calculations.

But the lull in growth isn’t expected to last. Dell’Oro Group tipped capex to surge 11% in 2024.

“New AI applications such as generative AI will be a key investment driver in the cloud and enterprise,” Dell’Oro Senior Research Director Baron Fung stated. “Furthermore, we anticipate demand for general-purpose servers to recover following a prolonged correction, and as customers make a transition to new server platforms that enable more efficient computing.”

Riding the wave

Growth beyond this year could be even more impressive. AMD CEO Lisa Su said last month the company believes the market for AI chips could hit $400 million by 2027. New Street Research’s Pierre Ferragu and team modeled out what that prediction would mean for AI-related data center capex overall (including spending on AI networking equipment, buildings, cooling, power and more).

The result? An AI-specific spend figure that jumps from $90-ish billion in 2023 to roughly $800 billion in 2027. That’s right, AI data center spending alone could nearly triple today’s overall capex figure.

By NSR’s calculations, that would put the overall data center capex figure for 2027 at $1.1 trillion. For the record that’s a 40% CAGR.

The team acknowledged that this scenario is a bit of a stretch, pointing out that the $1.1 trillion figure is triple current consensus estimates. Plus, they noted “global AI revenues [are] required to justify spending on AI chips growing to $400bn in 2027.”

However, NSR concluded “This nevertheless remains a very realistic scenario…We are so far on track for it, at the pace of current deployments, and nothing indicates a slowdown or an approaching bubble…Either it materializes in a healthy and sustainable way, or we fall short of this prediction, with disappointments, overinvestments, and low returns.”