Data Center REITs Face 2027 Workload Shift—Edge Play or Hyperscale Overhang?

The AI buildout is not a trend; it is a paradigm shift demanding exponential infrastructure investment. The scale is staggering. Hyperscalers alone are projected to spend $650 billion on AI infrastructure in 2026, a 71.1% year-over-year increase. This isn't just a bump in spending-it's the foundational capital required to power the next computing era. Barclays analysts see this as a multi-year inflection, identifying more than 400 public and private companies across 19 infrastructure subcategories that will benefit. The demand is so broad it spans from semiconductors to cooling systems, creating a vast ecosystem of winners.

For investors, the most direct vehicle into this S-curve is data center real estate. As AI hardware needs a home, the demand for massive, power-rich facilities is surging. Data center REITs like Digital Realty (DLR) and Equinix (EQIX) are positioned as key beneficiaries. These companies own the critical rails-massive facilities equipped with power, cooling, and connectivity-that hyperscalers and enterprises need to deploy their AI fleets. Their long-term leases provide stable, inflation-hedged revenue streams, making them a strategic play on the sustained, multi-year buildout of AI infrastructure.

The Physical Bottlenecks: Power, Cooling, and Construction

The AI buildout is hitting physical walls. While the financial and technological S-curves are clear, the real test for data center REITs is operational execution. The primary bottleneck is no longer just raw power, but speed to power. Hyperscalers need facilities online yesterday, but local grid constraints and community support are becoming critical site selection criteria. This creates a race against local infrastructure, where a willing population and a grid that can handle a surge are as important as the land itself.

Construction costs are rising to meet these demands. The global average cost for a new data center is forecast to increase 6% to $11.3 million per MW in 2026. This inflation in the build cost directly pressures REIT margins and project economics. It also favors companies with scale and operational efficiency, as the higher upfront capital required acts as a barrier to entry for smaller players. The sector is in an infrastructure investment supercycle, with roughly 100 GW of new capacity anticipated to come online between 2026 and 2030, a doubling of global capacity that will require trillions in real estate and tenant investment.

Cooling is the other critical infrastructure layer. As rack power requirements scale toward megawatt thresholds, air cooling hits its limits. This is driving a surge in demand for advanced solutions like liquid cooling. Companies like Eaton, Parker-Hannifin, and Vertiv are positioning as leaders in this space, with Vertiv even announcing a $50 million investment to expand its manufacturing facilities for cooling solutions. For REITs, this means the value of a facility is increasingly tied to its thermal infrastructure. A property with integrated liquid cooling capacity will command a premium lease and attract the most power-hungry AI tenants.

The bottom line is that the winners in this phase will be those who can navigate the physical constraints. They need sites with fast-track power connections, the capital to absorb rising construction costs, and the foresight to build in advanced cooling from day one. The REIT that masters this trifecta will capture the lion's share of value as the AI infrastructure layer is laid down.

Valuation and Risk: The REIT Layer and Market Dynamics

The investment case for data center REITs is built on a powerful, long-term S-curve. They own the physical infrastructure that will house the AI paradigm shift, providing stable, inflation-hedged revenue through long-term leases. This makes them a core infrastructure play. Yet, like any exponential bet, the path is not without friction. The sector faces clear financial and market risks that could pressure profitability.

The most direct headwind is interest rate sensitivity. As REITs are capital-intensive, their cost of debt is a major expense. In a higher-rate environment, refinancing and new development become more costly, squeezing margins. This is compounded by the risk of market oversupply. The projected build-out is massive: roughly 100 GW of new capacity is anticipated to come online between 2026 and 2030. While demand is strong, a misstep in timing or a slowdown in tenant adoption could lead to vacancy and lease rollover pressure, testing the sector's ability to maintain its premium pricing power.

Yet the scale of the opportunity dwarfs these near-term risks. The current power demand from AI is still modest, with AI representing roughly 0.6% of total US electricity in 2024. But this is a starting point, not a ceiling. The projected capacity expansion suggests this will become a major driver for the grid and energy sector, creating a new, massive load center. The sector's growth trajectory is set for the long haul, with the global data center sector projected to expand at a 14% CAGR through 2030.

The bottom line is one of managed tension. The REIT model provides a way to capture the value of this infrastructure supercycle with relative stability. The risks of interest rates and oversupply are real, but they are operational and cyclical, not existential. The fundamental demand driver-the exponential growth of AI and data-is structural and multi-year. For investors, the choice is between betting on the infrastructure layer itself, which offers a clear path to cash flow, or chasing the more volatile, high-growth tech companies that will occupy it. The REITs, with their scale and balance sheets, are positioned to be the foundational rails of the next computing era.

Catalysts and What to Watch

The thesis for data center REITs is clear: they own the rails for an exponential infrastructure build-out. The real test is in the near-term signals that will validate the pace and nature of this growth. Three key areas will determine whether the S-curve continues its steep ascent or faces a plateau.

First, watch for the structural shift in 2027. The demand profile is changing. Currently, AI only represented about a quarter of all data center workloads in 2025, with training driving the surge. But a significant inflection is anticipated in 2027, when inference workloads could overtake training as the dominant AI requirement. This matters because inference is a continuous, latency-sensitive task that demands a different infrastructure profile-more distributed, lower-latency facilities closer to end-users, not just massive centralized training clusters. REITs that have diversified their portfolios beyond hyperscale mega-facilities into edge and regional data centers will be best positioned to capture this new wave of demand.

Second, monitor the physical expansion hurdles. As the sector scales, the bottlenecks are shifting from pure power to local execution. The willingness of populations to support data centre growth is key, creating potential hurdles to expansion. This is coupled with long grid-connection queues in Europe and similar challenges elsewhere. For REITs, this means site selection is becoming a competitive moat. The companies that can secure fast-track power agreements and community buy-in will accelerate their development pipelines, while others face delays and higher costs. This is a structural shift where operational execution will separate leaders from followers.

Third, track the cost and cooling front. Construction costs are rising, with the global average forecast to increase 6% to $11.3 million per MW in 2026. This inflation pressures margins and project economics. More importantly, it accelerates the adoption of advanced cooling. As rack power scales, air cooling hits its limits. The evidence shows a clear investment push in this space, with companies like Vertiv investing $50 million to expand manufacturing for cooling solutions. For new builds, the adoption rate of liquid cooling will be a critical metric. Facilities that integrate this technology from day one will command premium leases and attract the most power-hungry tenants, directly impacting a REIT's asset value and rental income trajectory.

The bottom line is that the next phase of growth is about adaptation. The exponential demand curve is intact, but the winners will be those who navigate the 2027 workload shift, master local expansion hurdles, and build cooling-efficient infrastructure at scale. These are the catalysts that will turn the long-term thesis into tangible, outperforming results.