Data center investment has matured into a distinct segment of private markets. What began as a specialized real estate strategy is now shaped by power availability, contracted capacity, and long-duration operating performance. For allocators, the relevant exposure increasingly resembles infrastructure and contracted cash-flow assets rather than traditional property.

This shift matters because it changes how value is created, how risk is priced, and how capital should be governed over time.

Capacity as the Core Economic Unit

Global data center demand has accelerated beyond initial cloud consolidation into an era of massive AI-driven workloads. According to revised estimates from the International Energy Agency and industry analysts, global data center electricity consumption exceeded 460 terawatt-hours in 2024 and is projected to approach 1,000 terawatt-hours by the end of 2026. This 100% increase in two years is driven by the transition from general-purpose compute to high-density AI training and inference.

In the United States, data centers now account for approximately 5 percent of total electricity demand, up from 2 percent just five years ago. In primary markets like Northern Virginia and the “Silicon Prairie” (Ohio/Texas), data centers represent the single largest driver of new grid infrastructure.

As demand has scaled, the binding constraint has shifted away from physical shells toward usable megawatts, interconnection rights, and delivery timelines. Capital now follows capacity, specifically, “speed to power”, rather than square footage.

Contract Structures and Revenue Characteristics

Most institutional data center investments are anchored by long-term capacity contracts. These agreements typically span 10 to 15 years. However, in 2026, we see a bifurcation in how these are structured:

• Training Hubs: Massive, 500MW+ campuses with 15-year “take-or-pay” contracts, serving as the foundational infrastructure for Large Language Models.
• Inference Nodes: Distributed facilities closer to urban centers, often featuring shorter, more flexible 7–10 year terms to accommodate evolving edge-computing needs.

Contracted revenues display bond-like characteristics, but their stability depends on Pricing Design. Fixed-price contracts provide revenue visibility but embed margin risk. In 2026, sophisticated allocators insist on full power pass-throughs and CPI-linked escalators to protect against energy market volatility and rising labor costs for skilled technicians.

Valuation Inputs and Structural Sensitivities

Valuation in this segment reflects a combination of contracted cash flows and operational assumptions. Three factors now dominate outcomes:

1. Power Procurement and “Behind-the-Meter” Solutions

Electricity is the primary operating cost. Due to grid congestion, 2026 has seen a surge in Behind-the-Meter (BTM) generation. Top-tier operators are now integrating on-site natural gas turbines or Small Modular Reactors (SMRs) to bypass utility queues. Facilities with “self-generation” capabilities command a valuation premium due to decreased reliance on the public grid.

2. Grid Access and “Fast-Track” Interconnection

Interconnection delays remain a hurdle. In regions like PJM, new “Fast-Track” rules reward projects that are “shovel-ready” with proof of financing. Projects stuck in traditional queues face significant “dead equity” risk, where capital is deployed but cannot generate yield for years.

3. The “Social License” to Operate

A new valuation variable in 2026 is regulatory and community risk. Local opposition to water consumption for cooling and noise from high-density racks has led to project cancellations. Investors now treat permitting and environmental mitigation as core credit risks rather than administrative hurdles.

Common Sources of Mispricing

Several assumptions continue to distort underwriting in the space:

• Uniform Demand: Capacity is location-specific. A megawatt in a Tier 1 market with low latency is worth significantly more than a megawatt in a remote “cold storage” site.
• Infrastructure “Greenwashing”: Not all data centers are infrastructure. Facilities with high tenant churn or older power densities (below 20kW per rack) are increasingly viewed as “legacy real estate” and face accelerating obsolescence.
• Utility Capex: In 2026, many utilities require data centers to pay 100% of upfront grid expansion costs. Underwriting that fails to account for these “Line Extension” fees will significantly overstate Day 1 returns.

Portfolio Characteristics and Role

From a portfolio construction perspective, data center capacity offers long-duration cash flows with moderate correlation to public markets. While correlation is lower than traditional real estate, it is highly sensitive to energy price shocks and interest rate cycles.

Liquidity remains concentrated in secondary markets for stabilized assets. For allocators, the primary question is whether they are seeking development yield (taking power and construction risk) or core infrastructure yield (buying stabilized, power-secured cash flows).