Beyond data centers, the AI trade is a quality bet

Brian Klinksiek, global head of research and strategy at LaSalle, is making the case that AI's real estate payoff will arrive as a quality split inside property sectors, well before it shows up in headline rent data. The data center has become the visible address of AI in institutional real estate, the property type that answers where the computing boom physically lives. In an interview with IREI this week, he said allocators are reading that map too narrowly: AI's fingerprints are on every property type, and the effects will show up in both demand and property operations.

The backdrop is a market still working through higher-for-longer interest rates and a delayed recovery. LaSalle's mid-year update frames the moment around three AI-related tipping points: the data center investment boom, uncertain productivity gains, and a widening split between AI winners and losers inside real estate. The interview is partly a guide to telling genuine tipping points from compelling but unproven themes.

Klinksiek's sharpest distinction cuts across sectors rather than between them. On demand, he told IREI, sector-versus-sector comparisons will probably matter less than divides inside a sector, such as along quality lines, a shift that moves the allocator's question from which property type will benefit to which buildings inside each type will behave like AI-era assets.

That is a harder underwrite than the data center trade because a data center is a deliberate, purpose-built answer to AI demand, while the within-sector version requires picking winners before the purpose-built answer has arrived. Klinksiek's point that rent growth may be emerging before it shows up in headline data explains the mechanics: the spread between top-quality and secondary assets in the same sector should start to widen before any sector-wide average moves, so an allocator who waits for the index confirmation will be buying the theme after the premium is already in the price.

Underneath the sector talk sits the portfolio question. The interview asks whether real estate deserves a larger role in diversified portfolios, and the evidence Klinksiek lays out points toward better-selected real estate rather than simply more of it, because an uneven AI recovery makes the sector-level return a blend of winners and losers. The portfolio case then depends on being able to hold the winners, which turns the quality split into a mandate matter rather than a research observation.

The pattern is not confined to sectors that look like technology real estate. In logistics, occupiers' hunger for flexibility has made the length of the rent roll the number to watch, because the operating story inside the sector tells allocators more than top-line growth. Something similar is likely to happen across property types as AI upgrades the operating story of ordinary buildings: Klinksiek's point about property operations is that AI will change how buildings are run, and the winners will be owners who can integrate the technology into their own cost structure.

Meanwhile, the data center has become an increasingly specialized underwrite, one this publication has argued separates sponsors who can source power from those who cannot: an infrastructure bet wearing real estate's clothes that deserves its own allocation discipline. Klinksiek's contribution is to point allocators toward the next layer: the AI split that will play out inside ordinary portfolios, where the difference between a winning and losing building comes down to asset quality and the operating capability of its owner.

The immediate implication is practical, and it answers the deployment question at the center of the interview: where capital can go today despite the delayed recovery. Data centers have consumed the conversation and a large share of new allocation, while the within-sector quality trade is less visible and cheaper to enter. Rent growth emerging before it shows up in headline data amounts to an information advantage for allocators willing to underwrite buildings rather than sectors.

None of this is a call to ignore AI's most obvious property expression. The data center boom is real, and its construction pipeline will keep absorbing capital. But the allocators who extract durable returns from AI will be the ones who treat the data center as the starting point and build the quality screen into every sector they own. The next few rent reports will show whether the spread between top-quality and secondary assets starts to widen before the sector averages do.