The 3 Deficits of AI Infrastructure: Power, Water, and Land Shortages Driving Rent-Seeking [EN]

"The growth of the software industry cannot be free from the constraints of physical infrastructure. When the expansion of artificial intelligence models collides with limited resources—power, cooling water, and land—a structural change occurs where the relative economic value of infrastructure companies holding these resources increases."
— AI Infrastructure Analysis Framework (April 2026)

* The original data and baseline analysis of this macroeconomic shift are available in the Korean report. -> Korean Version

Prologue: Perspective of the Analysis

This report analyzes a key issue in the 2026 global macroeconomy: the phenomenon of AI infrastructure expansion colliding with physical constraints such as power grids, water resources, and land. The surging computing demand resulting from the rapid scale-up of AI models is clashing with the supply constraints of the physical infrastructure required to support it. As of [April 2026], delays in permits for new data center construction and regional imbalances in power supply and demand are emerging, and these structural changes are impacting asset allocation in capital markets. Investors need to examine how these physical constraints might impact the shifting economic value of data center REITs and utility companies.

EXECUTIVE SUMMARY

The AI infrastructure investment environment in 2026 is being shaped not only by technological innovation but also by physical constraints. The performance improvement of AI chips and the expansion of model sizes entail substantial power consumption and cooling water demand. Simultaneously, as the pace of national power grid expansions and new data center construction approvals fails to adequately meet this demand, the relative positioning of data center operators who have already secured power and land is likely to strengthen.

During this process, the business environments of data center REITs and utility companies are changing, and investors can consider the impact of these structural changes on their medium-to-long-term portfolios. However, this analysis does not recommend the purchase of specific assets, and it must be considered that outcomes may vary depending on the interaction of multiple macroeconomic variables.

01. Macroeconomy: Interaction Between AI Expansion and Physical Infrastructure

└ Increasing Trend in Data Center Power Density (Hard Data)

As the size of AI models increases, the power consumption density of data centers is also rising. According to the U.S. Department of Energy (DOE)'s 2024 Data Center Energy Usage Report, the power density of the latest high-performance computing facilities is showing significantly higher levels compared to legacy facilities. In CBRE's January 2026 North America Data Center Trends Report, data center rents in major hubs (such as Northern Virginia) are facing upward pressure, which is evaluated as reflecting the imbalance between supply constraints and increasing demand.

└ Regional Grid Constraints and New Construction Permitting Trends

According to data from the U.S. Energy Information Administration (EIA) and the Federal Energy Regulatory Commission (FERC), some regions require an expansion of power grid infrastructure to meet the additional power demand from new data center construction. Concurrently, due to environmental and infrastructure-related concerns from local residents and government authorities, the approval process for new data centers is showing regional disparities. Such regional deviations are creating differences in the competitive landscape between data center operating companies possessing existing facilities and new entrants.

02. [Risk Transfer Timeline] The Phased Impact of Physical Constraints on the Market

└ Sequential Impact of Physical Constraints on Asset Markets

These physical constraints can exert the following phased impacts on asset markets:
* Phase 1 (Recognition of Supply Constraints): As constraints on new data center construction become clear, the relative value of companies holding existing facilities is highlighted. [Price Variable] Adjustment of valuations (FFO multiples) for Data Center REITs.
* Phase 2 (Possibility of Rate Hikes): If supply constraints persist, conditions may form where existing data center operators can improve profitability through rent increases. [Price Variable] Changes in the operating profit margins of Data Center REITs and Utility companies.
* Phase 3 (Responses of Big Tech Companies): In response to rising data center costs, Big Tech companies may review various counter-strategies, such as expanding proprietary infrastructure construction (vertical integration), regional decentralization, or adjusting AI investment schedules. [Price Variable] Announcements of Capital Expenditure (CapEx) plans by Big Tech companies and the resulting market reactions.
* Phase 4 (Adjustment of Industrial Structure): Through this process, the value chain structure of the AI infrastructure industry may be adjusted, and the relative profitability distribution between tech companies and infrastructure companies may shift. [Price Variable] Relative stock price trends of tech stocks versus infrastructure companies.

03. System Architecture: The Relationship Between Big Tech Companies and Infrastructure

└ The Scope and Limits of Vertical Integration

Major cloud providers (AWS, Microsoft, Google) possess their own proprietary data centers, yet they cannot entirely control all physical resources. Securing land, obtaining grid interconnection rights, and acquiring water intake rights require regulations and approvals from local governments and existing utility companies. This structural characteristic implies that when Big Tech companies expand infrastructure, they are influenced by external factors (regulations, local policies, rates set by companies holding existing facilities).

└ Changing Bargaining Power of Infrastructure Companies

Data Center REITs like Equinix and Digital Realty mention rent increases and operational efficiency improvements in their 2026 10-K reports submitted to the SEC. Under supply-constrained conditions, already operational facilities may gain relatively strong negotiating positions, which can positively impact the companies' profitability and cash flows. However, whether these changes are sustainable and to what scale will depend on future market trends.

04. Capital Ecosystem Shifts: Rise in Relative Value of Infrastructure Assets

└ Increasing Investment Interest in Infrastructure Assets

As physical constraints become evident, private equity and infrastructure investment funds are showing interest in data center-related assets and land with power grid accessibility. This reflects the market's assessment that the medium-to-long-term trend of increasing AI infrastructure demand can elevate the economic value of real assets. McKinsey's (Dec 2025) report "The Future of Data Centers and AI Infrastructure" also notes that data centers and power infrastructure are critical constraining factors for the AI industry.

└ Re-evaluation of the Role of Energy and Water Utilities

The importance of utility companies that supply the electricity and cooling water essential for data center operations is also rising. According to a March 2026 report by the National Association of Regulatory Utility Commissioners (NARUC) and EIA data, electricity rates and supply costs in some regions are showing an upward trend, which can positively impact the profitability of utility companies. Reports by the World Resources Institute (Sept 2025) and the U.S. Geological Survey (March 2026) point out that water scarcity is a severe issue in certain regions, which can act as a constraining factor on data center operations.

05. Historical Comparative Analysis: Structural Impact of Energy Constraints

└ Comparison with Past Energy Crises

The 1970s oil energy crisis had widespread impacts across the economy, significantly elevating the economic value of companies that possessed or could supply resources. The fact that the relative scarcity of resources—power and cooling water—is emerging during the current AI infrastructure expansion process exhibits similar structural characteristics. However, there is a difference: the current situation is not a crisis where energy supply is completely cut off, but a supply-demand imbalance where the pace of supply growth fails to keep up with the pace of demand growth regionally.

└ The Role of Regulation and Environmental Policy

Whereas OPEC controlled the supply of physical resources during the past oil crisis, in the current scenario, the environmental and infrastructure policies and regulations of local governments are acting as the controls over new data center construction permits. This suggests that resource scarcity can be shaped not only by natural conditions but also by policy choices. It must be considered that the market structure could change significantly depending on future deregulation or tightening of regulations.

06. Potential for Easing Structural Constraints: Review of Alternative Scenarios

└ Q1. What if Small Modular Reactors (SMRs) or renewable energy are deployed in earnest?

[Analysis]: The commercialization of Small Modular Reactors (SMRs) and the expansion of renewable energy are primary pathways that could ease power supply constraints in the long term. According to the U.S. Department of Energy's (Nov 2025) SMR deployment timeline, commercial-scale operations are expected in the early to mid-2030s. During this interim period, power supply shortages may persist, but post-SMR commercialization, regional power constraints could be significantly alleviated. However, it must be considered that more than five years are required from the present (2026) to actual large-scale power supply.

└ Q2. What if data center technological efficiency improves drastically?

[Analysis]: Improvements in chip power efficiency are a positive factor that could reduce data center power consumption. However, as pointed out by the 'Jevons Paradox' in economics, improvements in individual chip efficiency do not necessarily lead to a decrease in overall system power demand. Companies that secure more efficient chips might utilize the resources to install more chips or train even larger models rather than saving power through efficiency gains. Therefore, chip efficiency improvements alone are expected to be insufficient to offset the absolute increase in data center power demand.

└ Q3. What if massive data centers are built in locations with low regional constraints?

[Analysis]: Building data centers in regions where vast expanses of cheap land can be secured may be attractive from a cost perspective. However, there are several constraints. First is the issue of cooling water supply. According to the World Resources Institute (Sept 2025) report, water scarcity is not limited to specific regions, and massive water usage may be restricted in some areas for environmental and social reasons. Second is the issue of data transmission latency. For applications where low latency is critical—such as financial trading or autonomous driving—the location of the data center is paramount, and remote placement could lead to decreased practicality. Third is local government regulation. Massive data center construction could be restricted depending on environmental and social impact assessments.

Macro Scenario: Probabilistic Future Trajectories

└ Scenario A (Base Case): Short-Term Persistence of Physical Constraints and Structural Adjustments (50%)

As physical constraints (power, land, water resources) fail to be resolved in the short term relative to the increasing demand for AI infrastructure, a phase persists where the relative positioning of companies holding existing facilities strengthens. There is a possibility that the profitability of Data Center REITs (Equinix, Digital Realty) and utility companies will improve, which could lead to a readjustment of their valuations. The S&P 500 Utilities sector (XLU) may show relatively strong performance compared to tech stocks. Big Tech companies may adjust their CapEx schedules or accelerate the construction of their own proprietary infrastructure.

└ Scenario B (Structural Shift Case): Policy Changes and Accelerated Infrastructure Expansion (35%)

Trigger: Governments in the U.S. and Europe aggressively push forward with deregulation policies for SMR commercialization or grid modernization.
Result: As permits for new data center construction expand and new supply enters the market, the relative superiority of existing facilities could weaken. Upward pressure on data center rents eases, and the extent of profitability improvements for infrastructure companies could be limited. In this case, there is a possibility that Big Tech companies will resume their CapEx expansions, and demand in the semiconductor and related industries could recover.

└ Scenario C (Tail Risk Case): Global Macro Shocks and a Plunge in AI Investment (15%)

Trigger: Due to global economic recession, surging interest rates, or escalating geopolitical risks, the profitability outlook for AI infrastructure investments sharply deteriorates.
Result: As Big Tech companies drastically reduce or postpone their CapEx plans, data center demand plummets. In this case, while the impact of physical constraints would temporarily ease, the profitability of infrastructure companies could actually worsen. In an industry-wide correction phase, there is a high likelihood that both tech stocks and infrastructure companies would be negatively impacted.

Implications from an Investor's Perspective

└ Portfolio Construction Considerations

To reflect the structural changes related to AI infrastructure in their portfolios, investors may review the following:
① [Relative Value of Infrastructure Companies]: Data Center REITs (rent trends verifiable in the 10-K reports of Equinix, Digital Realty) and Utility companies (S&P 500 Utilities ETF, XLU) have the potential to become structural beneficiaries of the increasing demand for AI infrastructure. However, this does not recommend the purchase of specific assets and can be considered according to the investor's own risk tolerance and portfolio goals.
② [Relative Valuation of Tech Stocks]: The CapEx plans and actual execution trends of AI chip and cloud service providers can be verified through upcoming earnings releases (10-Q, 10-K). How infrastructure constraints impact the growth strategies of these companies is a crucial monitoring indicator.
③ [Monitoring Macro Policy Changes]: The commercialization schedule for SMRs, grid modernization plans, and regulatory trends related to data centers are signals indicating turning points in the industrial structure. Such policy changes can significantly alter the market structure over time.

└ Risk Management Perspective

Points to note when considering AI infrastructure-related investments include:
① [Policy Risk]: Regulations related to data centers and power infrastructure are heavily influenced by local governments and environmental groups. Sudden shifts in regulatory direction can upend existing forecasts.
② [Technology Risk]: The commercialization schedules for new technologies (SMRs, renewable energy, chip efficiency) are difficult to predict, and the timeline for resolving physical constraints may vary accordingly.
③ [Demand Volatility Risk]: If the growth rate of the AI industry itself differs from expectations, infrastructure demand can shift in tandem. There is no guarantee that the current high growth expectations will persist.

Conclusion

The AI infrastructure industry in 2026 is experiencing a collision between the pace of technological innovation and physical constraints. Compared to the rapid scale-up of AI models, a structural supply-demand imbalance is emerging where the supply of basic resources like power, cooling water, and land fails to keep pace. This can act as a factor elevating the relative economic value of infrastructure companies. However, the duration of this structure will vary greatly depending on future policy changes, technological developments, and global macroeconomic trends. While recognizing these structural changes, investors must simultaneously review alternatives and risk factors that could overcome them. The current physical constraints could be a temporary issue or a long-term structural characteristic, necessitating continuous monitoring and verification to distinguish between the two.

※ Disclaimer

This report does not solicit the purchase or sale of any specific assets (including ETFs, REITs, and individual stocks), nor does it make value judgments on any specific industry. It is an article of macroeconomic analysis based on public data such as data center energy usage, grid constraints, and rent trends. Actual results due to future policy changes, technological developments, and rapid shifts in the global macroeconomy may differ from predictions, and the responsibility for all investment decisions lies with the investor. It is recommended to comprehensively review one's financial situation, risk tolerance, and investment horizon before making investment decisions, and to seek expert consultation if necessary.

Sources and References

[¹] U.S. Department of Energy (DOE) — Data Center Energy Usage Report — (Dec 2024, updated year-round) — https://www.energy.gov

[²] International Energy Agency (IEA) — Electricity 2024: Analysis and Forecast to 2026 — (Feb 2024) — https://www.iea.org/reports/electricity-2024

[³] U.S. Energy Information Administration (EIA) — Electricity Data and Grid Reliability Reports — (Quarterly updates) — https://www.eia.gov

[⁴] CBRE — North America Data Center Trends Report — (Jan 2026) — https://www.cbre.com/insights/reports/north-america-data-center-trends-2026

[⁵] JLL — Global Data Center Outlook — (Nov 2025) — https://www.jll.com/en/trends-and-insights/research/data-center-outlook

[⁶] Equinix, Inc. — Annual Report (Form 10-K) — (Feb 2026) — https://investor.equinix.com

[⁷] Digital Realty Trust, Inc. — Annual Report (Form 10-K) — (Feb 2026) — https://investor.digitalrealty.com

[⁸] National Renewable Energy Laboratory (NREL) — Data Center Load Growth and Grid Constraints — (Oct 2025) — https://www.nrel.gov

[⁹] World Resources Institute (WRI) — Water Use in Data Centers — (Sept 2025) — https://www.wri.org

[¹⁰] U.S. Environmental Protection Agency (EPA) — Water Rights, Cooling, and Industrial Water Use — (March 2026) — https://www.epa.gov

[¹¹] U.S. Geological Survey (USGS) — Water Availability and Regional Stress Indicators — (March 2026) — https://www.usgs.gov

[¹²] McKinsey & Company — The Future of Data Centers and AI Infrastructure — (Dec 2025) — https://www.mckinsey.com

[¹³] BloombergNEF — AI, Power Demand, and Grid Bottlenecks — (Jan 2026) — https://about.bnef.com

[¹⁴] U.S. Securities and Exchange Commission (SEC) — Form 10-K / 10-Q filings for data center REITs — (Updated year-round) — https://www.sec.gov

[¹⁵] U.S. Department of Energy (DOE) — Small Modular Reactors (SMR) Deployment Timeline — (Nov 2025) — https://www.energy.gov

[¹⁶] National Association of Regulatory Utility Commissioners (NARUC) — Grid Interconnection and Utility Rate Trends — (March 2026) — https://www.naruc.org

[¹⁷] Federal Energy Regulatory Commission (FERC) — Transmission and Interconnection Queue Data — (Real-time updates) — https://www.ferc.gov

[¹⁸] IEA — Artificial Intelligence and Electricity Demand — (Feb 2026) — https://www.iea.org

[¹⁹] CBRE Investment Management — Infrastructure and Data Center Land Banking — (Jan 2026) — https://www.cbreim.com

[²⁰] World Bank — Infrastructure, Land Use, and Energy Constraints in Digital Economies — (Oct 2025) — https://www.worldbank.org