The unprecedented artificial intelligence infrastructure building boom continues to come into focus, with the number of large-scale data centers in the U.S. set to jump by more than 200% in just four years.
:quality(75))
The U.S. has around 90 operational data center campuses with at least 100 megawatts of capacity, according to data published this week by capital markets analytics publication The Kobeissi Letter. But if all data center projects under construction or in earlier stages of development move forward, that figure could climb to 280 by the end of the decade — more than tripling the number of hyperscale facilities operating today.
That scale of build-out comes with an astounding price tag, according to The Kobeissi Letter. Over the same period, tech companies are expected to spend close to $7T on data centers, making the AI arms race one of the largest capital spending cycles in history.
Amazon, Microsoft, Alphabet and Meta have already signed more than $1.5T in data center leases since the advent of the AI infrastructure boom nearly three years ago.
This spending spree has expanded the data center industry well beyond its traditional hubs into a wide swath of communities across the country that have never seen this type of development. According to JLL, 77% of the data center construction pipeline is now moving forward in these "frontier markets."
There is a real possibility that many of the data centers in earlier stages of the development pipeline today don't end up coming to fruition. Growing community opposition to data centers has already derailed dozens of projects across the U.S., while ambiguous timelines for grid connections and uncertainty about the future of AI demand top a list of reasons planned data centers may not materialize.
Still, The Kobeissi Letter’s projections come on the heels of a report from PwC that also predicts a similar escalation in AI spending and data center build-out. PWC anticipates as much as $6.5T in global data center capital expenditures by 2030 and $31.6T through 2050, with a plausible upside of nearly $50T if AI adoption accelerates.
While The Kobeissi Letter's data suggests the pace of new data center construction could begin to plateau after 2030, PwC projects that investment in the sector will continue to climb. The report attributes this to the rapid refresh cycles for AI servers and other information technology equipment, which typically has to be replaced every four to six years. With each refresh, the capital cycle effectively starts again.
)
)
)
)
)
)
)
)
)
)
)
)