Mapletree data-center sale tests AI-era pricing

A 22-property Mapletree data-center portfolio has gone to market through JLL across 15 states, according to CoStar News, turning a book that has probably been carried as operating real estate into a pricing event for the AI-era market. The listing arrives amid what CoStar describes as soaring demand for facilities that support artificial intelligence, demand that has pulled data centers from the edges of institutional real estate into the center of allocation conversations. For a market chasing AI exposure, the test is whether the assets can be sold as one cleanly priced portfolio rather than a collection of one-off buildings.

The mandate is a liquidity test. Data-center capital has been moving at a pace that outruns the supply of power and land, and underwriting in the sector now looks more like infrastructure than property. The 15-state spread compounds the challenge: a buyer is pricing a national book of power markets, tax regimes, and lease structures that spans many metros, an underwrite closer to a utility portfolio than to a suburban office building. A portfolio shaped this way should draw yield-hungry pension funds and infrastructure investors if it can be priced cleanly. That Mapletree is testing the market now suggests the owner sees AI-driven demand as having pushed values to a point where a broad sale beats holding individual assets.

With no price, no target yield, and no buyer timeline disclosed, and no word on whether the portfolio will be sold whole or in pieces, the pricing question stays open. For private real estate capital, the sale will serve as a clearing-trade benchmark: if the portfolio trades at a cap rate that reflects infrastructure-style cash-flow durability, expect more institutional owners to bring data-center assets to market, and expect the bid to widen. If the auction stalls on the gap between traditional property underwriting and AI-driven growth expectations, the next several data-center portfolios may stay private longer.

Either outcome will shape the wider CRE capital cycle. Data centers are the asset class where the traditional real estate toolkit — rent roll, lease term, physical obsolescence — meets a demand curve shaped more like a utility build-out. Mapletree's mandate is an early visible test of how that collision prices at portfolio scale, and of the construction-cost squeeze that separates sponsors who can source power from those who cannot. The first cap-rate print will tell sponsors how far AI demand has moved the bid for data-center portfolios.