Richard Barkham says the property cycle completed in 2024 and 2025, without a crash

Half the capital I talk to is waiting for the data-center crash. The logic is respectable, even comforting. Every real-estate cycle ends the same way, the ending reprices everything, and the disciplined money buys powered land from the undisciplined money at sixty cents. All you have to do is wait.

Richard Barkham thinks you already missed the ending, because there wasn't one.

Barkham has been CBRE's global chief economist since 2014, sitting on top of CBRE Econometric Advisers and a research operation of roughly 700 people, which means his cycle calls are built on more property-level data than almost anyone else gets to see. His argument, laid out on a Nareit podcast in 2025, is that the traditional real-estate cycle completed in 2024 and 2025 without the usual crash. Rates rose, values corrected, the market digested it, and a new cycle is now underway, driven by macro fundamentals and, in his phrase, regional differentiation. No banking bust. No fire sale. The turn of the wheel happened while everyone was watching for the wreck.

Take him seriously for a minute, because the implication for data-center site selection is nasty. A crash is a repricing mechanism. It converts other people's entitlement mistakes into your discounted cost basis, and it forgives your own mistakes by resetting everyone's clock at once. If the cycle completed without one, that mechanism never ran. Nobody's mistakes got washed. Nobody's land came back to the market at sixty cents.

The discount arrives by place now, not by price.

Regional differentiation is a polite phrase for jurisdictions

When a macro economist says the new cycle will be driven by regional differentiation, the ground-level translation is that the spread between markets will do the work the crash used to do. And in this asset class, the spread between markets is not climate or labor or fiber. It is permission and power. Our two most instructive live files sit at the poles, and we score them 43 points apart.

Richland Parish, Louisiana, is the yes pole. On December 4, 2024, Governor Jeff Landry and Meta announced a $10 billion AI data-center campus on 2,250 acres, with an initial build-out of up to 4 million square feet. Louisiana Economic Development called it one of the largest private investments in state history. Site work began the same month. By mid-2025 Meta had bought roughly 1,400 additional acres. By late 2025 the total program, Meta's capital plus the generation and transmission build to serve it, had climbed past $27 billion, and in December 2025 Meta reported over $875 million in first-year contracts to Louisiana businesses.

Notice what the fight is about there, though. Not the campus. Entergy Louisiana petitioned the state's Public Service Commission for three new combined-cycle gas units totaling approximately 2,260 MW, part of a 5.2 GW generation filing, plus 240 miles of new 500-kV transmission, all framed as load-serving for the Meta site. The Sierra Club's Louisiana chapter and the Alliance for Affordable Energy intervened at the LPSC in mid-2025, and the live question in the docket is who pays: the single hyperscale customer the buildout serves, or every residential ratepayer in the territory. The generation package remains under review. Our read on the project is 78 out of 100, and the 22 points of hazard are almost entirely in that rate case, not in the parish.

South Memphis is the other pole. xAI bought a roughly 750,000-square-foot former Electrolux plant in June 2024 and moved at a speed the old cycle would have applauded. On-site gas turbines went in by late 2024, framed as temporary under a non-road-engine reading of the Clean Air Act. On July 17, 2025, an Inside Climate News investigation counted 27 to 35 turbines operating without major-source permits. The NAACP had already issued a notice of intent to sue that April. Shelby County's health department issued a gas permit in late 2025 over protest, the EPA closed the non-road-engine gap in January 2026, and on April 14, 2026, per CNBC, the NAACP, represented by the Southern Environmental Law Center, filed a Clean Air Act citizen suit in federal court seeking civil penalties and injunctive relief over the period of unpermitted operation. The machines run. The litigation runs too. We score it 35.

Here is what the old-cycle playbook would say about Memphis: so what, everyone cuts corners at the top, and the crash amnesties it. Barkham's point removes the amnesty. In a cycle that completes without a crash, there is no general repricing to bury a project-level mistake inside. The mistake stays attached to the asset, the docket stays open, and the precedent it sets follows the operator to the next site.

What you are actually waiting for

If Barkham is right, the thing my inbox is waiting for is not late. It is not coming. The correction already happened, quietly, in 2024 and 2025, and the new cycle's winners are being selected right now by a sorting mechanism that looks nothing like distress. It looks like a parish that said yes and a rate case that will decide the true cost of that yes. It looks like a federal docket and an EPA rule change deciding how expensive speed turns out to have been.

That sorting is legible today, in public records, before the spread shows up in pricing.

My bet, stated so you can check it: through the end of 2028 there will be no broad distressed repricing of powered data-center land in the United States. The spread that moves instead is the premium for sites holding both entitlement and interconnection over sites holding neither, and it widens every quarter that the AI demand curve holds. If a genuine fire sale in powered land shows up before then, Barkham was wrong and so was I, and averaging into it will be the trade of the decade.

But if you are running a site-selection process on a wait-for-the-bottom thesis, look at what 43 scoring points between Richland Parish and South Memphis actually represent, and ask your team one question. Which pole is your next filing closer to?