The Age of Strategic Assets: Energy, Data Centers, and the Geopolitics of Infrastructure Real Estate
Executive summary
The next cycle in U.S. commercial real estate is being shaped less by tenant mix and more by power, compute, and permitting. Data centers and adjacent “infrastructure real estate” (powered land, transmission‑proximate industrial sites, gas interconnects, and firm generation) are rapidly becoming strategic assets in the same sense as ports, pipelines, and logistics chokepoints. This is not a metaphor: the U.S. government’s own demand outlook now frames data centers as a material driver of national electricity consumption, with U.S. data centers rising from ~58 TWh in 2014 to ~176 TWh in 2023, and projected to reach ~325–580 TWh by 2028 (roughly 6.7%–12% of total U.S. electricity).
Three forces are converging. First, AI adoption is increasing compute intensity, pushing higher rack densities, more aggressive cooling requirements, and a stronger premium for “deliverable megawatts” over generic square footage. Second, grid constraints are tightening: in the U.S. interconnection pipeline alone, ~10,300 projects were seeking interconnection as of end‑2024 (representing ~1,400 GW of generation and ~890 GW of storage), signaling a system under structural queue pressure. Third, policymakers are treating infrastructure siting and control as a national security issue, expanding federal review of certain real estate transactions near sensitive installations and increasing scrutiny of ownership and proximity risks.
For investors, the implication is straightforward but underappreciated: infrastructure real estate is turning into an asset class defined by energy security - with a growing wedge between (a) assets that can secure power inside an 18–36 month window and (b) assets that cannot. In leading U.S. markets, cap rate frameworks are adapting: CBRE notes cap rates for Class A data center facilities commonly pricing ~100–150 bps above the 10‑year Treasury, while the scarcity of near‑term power access is raising land and development basis and pushing rent growth expectations above inflation.
Macro thesis
The core thesis is that “strategic assets” in real estate are broadening from traditional hard infrastructure (ports, rail, and energy transport) to compute‑anchored infrastructure: land and buildings whose value is determined by the ability to convert electricity into contracted digital services. This shift matters because AI workloads are not simply another form of office demand; they are a new industrial load category that competes directly with manufacturing electrification, residential growth, and energy transition requirements for scarce transmission capacity.
This is why the sector’s geopolitics feels different than prior tech‑driven CRE waves. In the last decade, hyperscale buildouts often optimized for tax regimes, fiber routes, and construction cost. In the next decade, the binding constraint is increasingly the deliverability of firm power and the political feasibility of adding generation and transmission fast enough to match AI deployment timelines. The International Energy Agency notes a key timing mismatch: a data center can be operational in two to three years, while energy infrastructure frequently requires longer lead times due to planning, approvals, and capital intensity.
At the corporate level, data center operators and hyperscalers are signaling that power availability and volatility are not peripheral risks; they are central operating variables. For example, Equinix highlights risk exposure to increased power procurement costs, outages, shortages, and capacity constraints, and notes that limitations in generation, transmission, and distribution can delay or restrict expansion in new and existing markets. Digital Realty similarly describes electric power as a significant component of operating expenses and points out that increased regulation or taxation of energy use could raise costs that may not be fully pass‑through to customers.
The geopolitical layer is that these assets sit at the intersection of energy policy, cyber resilience, foreign investment screening, and industrial competitiveness. The U.S. Treasury expanded the scope of federal review for certain real estate transactions near more than 60 military installations across 30 states, explicitly connecting land control to intelligence and defense concerns. In other words, infrastructure real estate is increasingly “regulated terrain,” not just leased space.
Quantifying AI compute demand and energy intensity
The quantitative takeaway is not merely “data centers use more electricity,” but that the rate of change is compressing decision cycles for utilities, regulators, and capital allocators. The U.S. Department of Energy (DOE) highlights a national step‑change: data centers represented ~4.4% of U.S. electricity in 2023 and are expected to represent ~6.7%–12% by 2028, implying rapid growth over a single development horizon.
Globally, the IEA estimates data center electricity consumption at ~415 TWh in 2024 (about 1.5% of global electricity) and projects consumption to ~945 TWh by 2030 in its Base Case, growing ~15% per year from 2024–2030: more than four times faster than total electricity growth in other sectors. The IEA also stresses that AI‑linked “accelerated servers” are a major contributor to incremental demand, and that growth concentrates in specific locations - creating localized grid stress even when national percentages appear manageable.
The range of outcomes remains wide. Electric Power Research Institute projects U.S. data centers could consume ~9%–17% of U.S. electricity by 2030, reflecting uncertainty in AI adoption rates, hardware efficiency, and buildout constraints. Importantly, this range is not purely “demand driven.” It is also a function of supply bottlenecks - especially power delivery timelines, interconnection studies, and transformer and switchgear availability.
AI intensity shows up physically as power density and cooling requirements. In the Uptime Institute Global Data Center Survey 2024, average rack densities were still reported as below 8 kW, and most facilities did not have racks above 30 kW (and those that did typically had only a few). Yet the distribution is shifting: Uptime’s 2025 survey commentary points to rising adoption of racks in the 10–30 kW range and emphasizes worsening power constraints and the need to modernize for higher density requirements. Cooling is the second‑order constraint: Uptime’s Cooling Systems Survey 2025 shows direct liquid cooling adoption remains gradual, but high rack density persists as the primary driver for adoption, alongside electricity costs and sustainability pressures.
This physical reality feeds directly into real estate underwriting. Higher density typically means more robust electrical rooms, greater redundancy, larger mechanical footprints, and faster obsolescence risk for older shells that cannot support heavier racks or modern distribution. It also changes the “best location” logic: inference workloads can favor proximity to end users for latency, while training can favor places with the largest contiguous blocks of power and land. CBRE explicitly notes that inference is reshaping demand toward more regional and distributed capacity, even as training clusters drive mega‑campus announcements.
Site selection and regional market dynamics
Across leading U.S. markets, site selection has become a multi‑variable optimization where “power deliverability” is the gating item and other real estate attributes are secondary. The practical screen now starts with (1) credible power availability within an investable timeframe, (2) utility and transmission constraints, (3) entitlement and community acceptance, and then (4) land, fiber, water, and tax considerations. CBRE’s national view underscores that many planned projects remain delayed due to permitting, zoning, and power procurement hurdles, even as vacancy in primary markets hits record lows.
In what follows, energy pricing should be interpreted cautiously: for comparability, state‑level average retail electricity prices can provide a directional proxy, but large users often negotiate tariffs, demand charges, and bespoke service agreements that diverge substantially from state averages.
Texas (Dallas–Ft. Worth / ERCOT): Dallas–Ft. Worth has become a primary beneficiary of hyperscale overflow and “power‑first” siting. CBRE reports total inventory of roughly 1,007 MW, vacancy around 2.4%, and ~701 MW under construction with ~94.5% preleased, alongside roughly 3 GW planned—a scale that indicates deep demand but also increasing competition for substations, transformers, and transmission upgrades. The market’s attraction includes a generally business‑friendly development environment and a large footprint of developable land; however, CBRE also notes that deliverable power is being throttled, with utilities tightening controls over entry into the “power queue,” reinforcing that the real scarcity is time to power. On energy costs, the U.S. Energy Information Administration’s December 2025 average retail price benchmarks show Texas commercial electricity around 9.03¢/kWh and industrial around 6.67¢/kWh (directional only for hyperscale underwriting). A representative “major transaction” illustrating the market’s scale is the PowerHouse Irving land acquisition: a 50‑acre purchase in Irving–Las Colinas planned for a multi‑phase campus totaling roughly 200 MW at buildout, reflecting the market’s tilt toward large powered shells and campus configurations.
Virginia (Northern Virginia / PJM): Northern Virginia remains the global benchmark for data center clustering, but its success is tightening constraints. CBRE reports total inventory of roughly 5,836 MW, vacancy near 0.5%, and a construction pipeline near 1,784 MW (with ~82.5% preleased), showing both continued dominance and limited “available now” options. The key siting challenge has shifted to power delivery timelines: CBRE notes that utility batching and extended delivery timelines have become a material underwriting variable as demand outpaces grid expansion. Permitting is increasingly political; CBRE highlights that community sentiment and local processes, including special exception requirements in parts of the region, can influence timelines even when demand is strong. The pricing pressure is visible not only in rents but in dirt: CBRE cites recent and pending transactions in Northern Virginia and the Northeast exceeding $8 million per acre—a signal that “powered land” is being capitalized like a scarce utility entitlement rather than conventional industrial land. On electricity benchmarks, Virginia’s December 2025 average retail prices were about 10.21¢/kWh commercial and 9.35¢/kWh industrial. In this market, “major transactions” often present as large site assemblages and new campus announcements rather than stabilized sales, with preleasing and off‑market deals substituting for traditional vacancy‑driven leasing.
Arizona (Phoenix metro): Phoenix has emerged as a key relief valve for West Coast demand and as a growth market for both hyperscale and colocation, but the region illustrates how fast power constraints can appear. CBRE reports that colocation space and power absorption totaled 256 MW in H2 2025, with multiple full‑scale projects expected to complete within the next 12 months across Mesa, Phoenix, Avondale, Goodyear, and Glendale. Critically, CBRE also flags that power‑delivery restrictions at major utilities (SRP and APS) are limiting availability of powered land, and that SRP is processing a first cluster study of 25 applicants - evidence of a market moving from “site selection” to “grid triage.” Average retail electricity price proxies in December 2025 were 11.87¢/kWh commercial and 7.22¢/kWh industrial in Arizona. A representative recent land transaction reported locally was a 70‑acre purchase in Glendale for $58,470,283 (Dec 2025), underscoring both the scale of land plays and the importance of large, infrastructure‑ready industrial corridors such as Loop 303.
The power stack and grid constraints
The “infrastructure real estate” stack supporting AI compute has four layers: generation (where energy is produced), transmission (how bulk power moves), distribution/substations (how power reaches the site), and the on‑site electrical and cooling plant (how power becomes usable compute). Each layer has distinct timelines, permitting regimes, and risk owners - making this a capital markets problem as much as a development problem.
Natural gas, nuclear, renewables, and hybrid power strategies. Over the next five years, the most investable reality for many U.S. markets is that incremental load growth will be met through a mix of grid purchases and local generation additions, with natural gas remaining a near‑term swing resource in many regions as demand accelerates. The EIA forecasts U.S. electricity load growth at 1.9% in 2026 and 2.5% in 2027, with the highest growth concentrated in ERCOT and PJM - exactly where hyperscale activity is densest. In that context, “energy security” is not only geopolitical; it is operational.
Nuclear has re‑entered the discussion as a firm, carbon‑free option for large loads. The EIA documents that data center owners have pursued dedicated nuclear supply arrangements, including a contract for 960 MW of capacity from a nuclear facility in Pennsylvania connected to PJM. Corporate statements reinforce this direction: Google and Kairos Power signed an agreement described as a pathway to deploy 500 MW of advanced nuclear projects by 2035. Microsoft and **Constellation Energy announced a power purchase agreement intended to support the return of a nuclear unit to service, adding approximately 835 MW of carbon‑free energy to the grid - illustrating how hyperscale demand can increasingly anchor generation economics.
At the same time, the underwriting playbook is changing: CBRE observes that grid power for many projects is “largely booked through 2030” in most major markets, contributing to the rise of on‑site power solutions (including natural gas generators and other behind‑the‑meter approaches) and emphasizing the importance of clear permitting for such configurations.
Transmission and interconnection as the binding constraint. The U.S. grid’s backlog is visible in the interconnection queue. Lawrence Berkeley National Laboratory reports that as of end‑2024, ~10,300 projects were seeking interconnection, and only a fraction of projects historically reach commercial operation. From a real estate perspective, this means the “effective scarcity” is not land - it is transmission capacity, substation availability, and deliverable interconnection rights within a defined timeline.
Policy is responding, but timelines remain long. New high‑voltage transmission can take about 10 years on average to complete, which is misaligned with hyperscale development cycles. Federal permitting reforms are attempting to compress portions of the process: DOE’s Coordinated Interagency Transmission Authorizations and Permits program establishes DOE as a coordinating lead and sets a two‑year deadline framework for qualifying projects. Meanwhile, Federal Energy Regulatory Commission Order No. 1920 and subsequent clarifications push transmission providers toward long‑term regional planning over a 20‑year horizon and more structured cost allocation-important, but not an immediate solution for near‑term power delivery in constrained hubs. Reliability authorities are increasingly explicit about large load risks: North American Electric Reliability Corporation warns that resource adequacy risks are intensifying as demand grows, and it has published frameworks addressing planning and operational risks from emerging large loads.
Typical development timeline in narrative form. A simplified “infrastructure real estate” timeline typically begins with site origination and control (often 2–6 months for market sourcing, exclusivity, and initial feasibility), followed by utility engagement and load studies (commonly 6–18 months depending on the market and whether cluster or batch studies apply), and then entitlements and local permitting (often 6–18 months, but longer where special exceptions, community opposition, or water constraints arise). Equipment procurement and construction for a modern multi‑building campus frequently runs 18–30 months (with commissioning and customer acceptance testing adding another 2–6 months), while any required upstream transmission expansion or major substation work can extend the critical path into multi‑year territory. This mismatch is why the IEA notes data centers can be built in roughly two to three years, while the broader energy system often requires longer lead times, and why investors now underwrite “time to power” as a central risk factor rather than an engineering footnote.
Capital markets implications and asset class winners and losers
Capital markets are repricing three things simultaneously: (1) the durability of data center cash flows, (2) the scarcity value of power access, and (3) the embedded optionality in land and infrastructure that can be converted into deliverable megawatts. CBRE reports record low vacancy in primary markets and rising asking rates, noting the average asking rate for a 250–500 kW requirement reached a record level in 2025, and that supply constraints are pushing preleasing and off‑market activity. **JLL similarly characterizes vacancy in North America as around 1% at year‑end 2025, reinforcing the view that sector fundamentals remain tight despite high construction levels.
Valuation mechanics are evolving. CBRE notes that cap rates for Class A facilities typically price at roughly 100–150 bps above the 10‑year Treasury, and that rent growth is expected to outpace inflation over the next two to five years due to rising land, construction, and power delivery costs. The “spread” is doing more work than in prior cycles: it must underwrite tenant credit, technology obsolescence, and power procurement risk, while still recognizing the scarcity of stabilized, powered product.
Credit markets are also adapting. CBRE highlights the growth of structured financing in the sector, including record issuance of single‑asset single‑borrower CMBS and large AI‑factory financing packages, while noting that power delivery and entitlement challenges can extend transaction timelines and reduce annual investment volume even amid strong demand. This points to a market where capital is available, but execution risk and delivery timing increasingly determine who earns development premiums.
The “winners and losers” are therefore better understood by power position than by traditional property sector labels. Likely winners include (a) stabilized data centers and powered shells with contracted capacity in constrained markets, (b) powered land with credible substation access inside a two to three year window, (c) industrial sites adjacent to substations, gas infrastructure, or transmission upgrades that shorten the time‑to‑compute, and (d) certain utility, generation‑adjacent, and grid‑enhancing infrastructure corridors that directly monetize congestion relief and capacity expansion. Likely losers include (a) “cheap land” without power deliverability, (b) older data center shells unable to retrofit for higher densities or cooling, and (c) markets where permitting or community constraints prevent load growth even when demand exists.
Investment strategy, due diligence, and five-year outlook
Investment strategy in this regime should be framed as infrastructure underwriting with real estate execution. The first strategic question is not “what rent can I get,” but “what is my credible path to power, and who bears the time and cost risk along that path.” CBRE’s observation that sites offering power access within 18–36 months are highly sought after provides a practical screen for where land basis can be justified and where it is speculative.
In portfolio construction, investors are increasingly forced to choose between three postures. One posture is “own the stabilized box,” emphasizing contracted cash flow and tenant underwriting, and accepting lower but durable yields. The second posture is “own the power optionality,” acquiring land and pre‑entitlements where power can plausibly be delivered, recognizing that much of the return is embedded in power scarcity rather than structural rent. The third posture is “own the energy interface,” investing in assets that monetize the translation between grid supply and compute demand (substation‑proximate land, behind‑the‑meter generation siting, and potentially co‑located energy and compute campuses where legally feasible).
A due diligence checklist, written as prose because it must be applied case‑by‑case, should begin with power reality: obtain utility‑verified deliverability milestones, not marketing estimates; confirm whether the site is subject to batch or cluster studies and what triggers re‑study; evaluate transformer lead times and any upstream network upgrade requirements; and map the site’s exposure to transmission congestion and curtailment risks. Next, validate entitlement path and political risk: confirm zoning and any special exception requirements, assess community sentiment, water constraints, and the local tax base’s dependence on data center revenue (which can cut both ways in approvals). Then, underwrite operating cost pass‑through and volatility: many operators explicitly flag power costs as material and not always fully pass‑through, so lease structure, tariff design, and hedging capability should be examined as carefully as credit. Finally, evaluate “geopolitical compliance”: assess whether the asset’s location creates foreign investment review sensitivity, and whether the partnership structure, tenant profile, or capital sources could trigger review or restrictions in a tightening national security environment.
Policy risk is no longer abstract. Federal scrutiny over sensitive location real estate has expanded through CFIUS rulemaking, and certain states are tightening rules around foreign participation in real property ownership and long‑duration leases. For example, Texas enacted restrictions effective September 1, 2025 targeting certain foreign ownership and long leases, which can affect deal structuring and exit liquidity for particular investor categories. Cyber and operational security should also be treated as “real estate diligence,” not IT diligence, because data centers are physical nodes in critical infrastructure and cyber guidance frameworks are evolving (including updates to NIST’s Cybersecurity Framework).
Five-year scenario analysis (2026–2031). In a base case, DOE’s 2028 range becomes the anchor: U.S. data center electricity use rises meaningfully toward the midpoint of the 325–580 TWh range by 2028, with continued tight vacancy and rent resilience in core markets, but with expansion shifting toward secondary and tertiary nodes that can deliver power faster. Capital continues to migrate into powered shells, structured financings, and utility‑adjacent land, with cap rates remaining relatively supported because scarcity and contractual cash flow offset higher rates. In an upside case, accelerated AI adoption and faster power solutions (including more rapid deployment of grid‑enhancing technologies, higher success rates in permitting reform, and effective long‑term regional planning under Order 1920) enable faster delivery of new capacity, supporting higher development volume and broader geographic dispersion as inference pushes regional buildouts. In a downside case, interconnection bottlenecks, community opposition, and delayed transmission projects constrain deliverability, pushing more developers toward behind‑the‑meter solutions and heightening policy friction around emissions, water, and local ratepayer impacts; this case raises execution risk premiums, widens the gap between “paper MW” and deliverable capacity, and could strand some land plays that were underwritten on optimistic time‑to‑power assumptions.
Sources and Disclaimer
Sources:
U.S. DOE summary of LBNL data center electricity demand outlook (2014–2028).
IEA “Energy and AI” analysis on data center electricity demand (2024–2030 base case and sensitivity cases).
EPRI “Powering Intelligence” projections for U.S. data center electricity share by 2030 (2026 edition).
Uptime Institute Global Data Center Survey (density and constraint themes; 2024 report and 2025 summary resources).
U.S. EIA (retail electricity price benchmarks; load growth outlook for 2026–2027).
LBNL “Queued Up” interconnection queue statistics (end‑2024 snapshot).
FERC Order No. 1920 and clarifications (long‑term regional transmission planning and cost allocation).
DOE CITAP program overview (two‑year federal permitting coordination framework for qualifying transmission).
NERC large load reliability risk framing and planning considerations.
CBRE North America Data Center Trends H2 2025 and market profiles (Dallas, Northern Virginia, Phoenix; valuations and power access commentary).
JLL North America Data Center Report (year‑end 2025).
Selected corporate disclosures and announcements related to power and expansion (Equinix risk factors; Digital Realty operating cost exposure; Google‑Kairos and Microsoft‑Constellation power announcements).
U.S. Treasury CFIUS real estate jurisdiction expansion (final rule context).
Texas foreign ownership restriction summary (effective date and scope context).
Disclaimer: This publication is for informational and educational purposes only and is not, and should not be construed as, investment advice, an offer to sell, or a solicitation of an offer to buy any security, product, or service. Any views expressed reflect judgment as of the date of publication and are subject to change without notice. Sterling Asset Group makes no representation or warranty, express or implied, as to the accuracy, completeness, or reliability of the information contained herein, which may be based on public sources believed to be reliable but not independently verified in all cases. Any forward‑looking statements, projections, or scenario analyses are inherently uncertain; actual outcomes may differ materially. Investors should conduct their own independent analysis and consult their own legal, tax, accounting, and financial advisers before making investment decisions.