Why data centre landlords are missing out on the AI boom — and how they can catch up
The AI buildout is driving huge demand for compute, but public data centre REITs have fallen behind. Payout rules, risk limits and power and land constraints explain the gap.
Copy link
By Torontoer Staff
The surge in AI-driven demand for computing power has created trillions in market value for chipmakers and cloud giants, and it has prompted plans for massive new data centre capacity. Yet the largest public data centre landlords have lagged in public markets. Equinix, Digital Realty and Iron Mountain have seen share prices fall over the past year even as AI spending accelerates.
The disconnect comes down to capital structure and risk appetite, along with operational limits such as grid capacity and available land. Those factors have made it harder for REITs to compete with better-funded private developers and deep-pocketed tech tenants that are building or pre-leasing huge AI facilities.
Why REITs are trailing the AI gold rush
Real estate investment trusts were built to return steady income to investors. By law they must distribute roughly 90 percent of taxable income as dividends. That model delivers predictable payouts, but it leaves little retained capital to fund the sort of multibillion-dollar, front-loaded investments AI infrastructure demands.
As a REIT, you can’t stick your neck out there in the way others are doing. It’s really held them back in a big way.
Andy Cvengros, co-lead, US data centre markets team, JLL
- Dividend rules limit retained earnings for new builds and speculative commitments.
- Conservative leverage policies reduce borrowing capacity compared with private competitors.
- Shareholders of public REITs tend to favour steady returns over high-risk, high-return bets.
- Operational risks such as project delays and contract termination clauses increase downside.
Competition, capital and changing deal dynamics
Big tech firms are funding their own builds or partnering with private developers and investors. Private operators can and do run far higher leverage, and they are backed by investment firms willing to take on construction and timing risk. That has shifted many of the biggest 2025 deals away from public REITs.
The market believes that chips from Google, Broadcom, Nvidia and others will capture the economic profits of AI, not mercenary data centre developers like REITs.
Mark Giarelli, analyst, Morningstar
Bank of America analysts note that the top data centre development contracts in 2025 were often won by non-REIT operators. Private players and vertically integrated firms can make large, speculative commitments that public REITs find difficult to match given shareholder expectations and balance-sheet limits.
Operational constraints: power, timing and location
Even when the capital exists, physical limits bite. Grid capacity is a bottleneck in many markets, and delivering large megawatt loads can be delayed by permitting, upgrades and equipment timelines. JLL reported that more than half of data centre projects in 2025 experienced delays of three months or more.
Contracts for big tenants can include early termination clauses tied to construction timing or other contingencies. That adds revenue risk for landlords who commit capacity up front. For cautious REIT investors, the combination of timing risk and potential stranded shells looks unattractive.
Where public landlords still have advantages
Demand for AI workloads is shifting from model training to model inference and delivery. That increases the value of low-latency sites close to users and well connected to fibre networks. Public REITs that own urban-edge facilities, including dense fibre and peering, are positioned to capture those workloads.
Analysts have pointed to Equinix as an example. Its focus on speed and connectivity, and its xScale business that partners with hyperscalers, align with the move toward distributed, latency-sensitive compute. Other firms that moved aggressively have seen sharp gains. Applied Digital pivoted from crypto hosting to AI customers and raised large private capital, and its market value and revenue have climbed rapidly.
What it will take to catch up
Public REITs have a limited set of levers. They can pursue joint ventures with strategic partners, sell or monetise non-core assets to free cash, or seek equity and debt raises targeted at AI builds. They can also lean into their core strengths, such as dense connectivity, and market those assets to inference workloads that need proximity to users.
To close the gap, REITs will need to accept a different balance of risk and return. That means hedging big bets carefully, securing long-term commitments from tenants before committing capital, and being prepared to deploy equity or partnered financing for large projects.
There is room to win. The AI buildout is vast and varied, and demand will not be met by a single model. Public landlords that adapt capital strategies, deepen partnerships with hyperscalers and prioritise connectivity can still capture meaningful share of the market.
For investors, the question is whether REIT management teams can move beyond predictable dividend-focused strategies to a model that balances income with selective, well-structured growth. For the operators, the choice is between sticking to core markets and cash flows, or taking on bigger, more speculative plays to keep pace.
AIData centresREITsReal estateEquinix
