Artificial intelligence does not run on software alone. Every model, application, and digital service depends on physical infrastructure, and every data center depends first on reliable electricity.

This article is the second post in our eight-part AI investment series. The series overview introduced AI as a connected economic value chain rather than a narrow group of technology stocks. The sequence begins with Raw Materials → Power → Physical Infrastructure → AI Semiconductor Supply Chain → AI Compute & Networking → AI Intelligence → AI Applications → Physical AI.

Power is the second layer. It sits between the materials required to build the system and the physical infrastructure required to house it. If electricity generation, transmission, and grid interconnection cannot keep pace, the rest of the AI value chain will encounter a practical constraint.

AI Data Centers Are Electricity-Intensive Facilities

A large AI data center can consume as much electricity as a small city. Unlike many commercial facilities, it often operates continuously, with high power density and demanding reliability requirements. Training and inference workloads place substantial demands on servers, networking equipment, cooling systems, backup power, and electrical distribution.

The International Energy Agency expects global data-center electricity consumption to approach roughly 1,000 terawatt-hours by 2030 in its base-case outlook, approximately double the level earlier in the decade. Exact forecasts vary because analysts must estimate future AI adoption, model efficiency, server utilization, and data-center construction. The direction of travel is clear: data-center load is growing rapidly, and AI is a primary driver.

The challenge is not simply producing more electricity. A data center needs power:

• In the right location
• At the right voltage
• With sufficient reliability
• On a timeline that matches the facility’s construction schedule
• At a cost that supports the economics of the project

That distinction matters. A region may have adequate generation capacity in aggregate but still lack the transmission, substations, or distribution equipment needed to serve a new hyperscale campus. The bottleneck can therefore exist at the grid-connection level even when total national generation appears sufficient.

The IEA’s analysis of energy demand from AI provides useful context on the relationship between computing growth and electricity consumption. For investors, the important conclusion is that AI expansion creates demand across the entire power system, not only for power producers.

Three Demand Drivers Are Reinforcing One Another

1. Data-center load growth

AI data centers are increasing electricity demand in concentrated geographic markets. This concentration can magnify the impact on local utilities and regional transmission organizations. A single large campus can represent a material new customer for a utility, while several campuses in the same region can require new generation, substations, transmission lines, and gas infrastructure.

Utilities and grid operators must also account for the timing and reliability of the load. A data center cannot easily reduce consumption during a heat wave, cold snap, or period of grid stress. That raises the value of firm capacity and well-planned system reserves.

2. Broader electrification

AI is arriving alongside other sources of electricity demand growth, including electric vehicles, heat pumps, industrial reshoring, hydrogen-related projects, and manufacturing facilities. These trends compete for the same generation and transmission resources.

As electrification expands, demand growth becomes less dependent on traditional residential and commercial consumption. The result is a more capital-intensive power system, with larger requirements for generation, grid modernization, and long-duration planning.

3. Corporate clean-power procurement

Many hyperscalers have public carbon-reduction objectives and are pursuing long-term power purchase agreements, renewable-energy contracts, nuclear arrangements, and other forms of dedicated supply. These commitments can support new generation, but they do not eliminate the need for a functional grid.

Wind and solar resources may be located far from data-center clusters. Transmission is required to move that power, and intermittent generation generally needs storage, firming resources, or broader market coordination. Nuclear power can provide carbon-free baseload generation, but new nuclear facilities require significant lead time.

Meanwhile, grid interconnection queues remain lengthy in many markets. Projects can spend years waiting for studies, approvals, transmission upgrades, and cost-allocation decisions. This creates a meaningful difference between announced power capacity and power that can actually be delivered.

Where Investors May Find Exposure

The power layer contains several distinct investment themes. They should not be treated as interchangeable, and each has a different risk profile.

Nuclear generation and uranium fuel

Nuclear power offers high-capacity-factor generation with minimal direct carbon emissions during operation. Existing nuclear fleets may benefit from license extensions, uprates, improved operating economics, and renewed demand for firm power.

The investment universe includes:

• Owners and operators of existing nuclear plants
• Uranium miners and fuel-cycle companies
• Engineering and equipment suppliers
• Developers of small modular reactors
• Companies pursuing long-term nuclear power agreements

Representative publicly traded examples include Constellation Energy, Cameco, Centrus Energy, and NuScale Power. Nuclear-focused ETFs such as the VanEck Uranium and Nuclear Energy ETF (NLR) and the Global X Uranium ETF (URA) provide different forms of exposure.

The distinction between existing nuclear assets and early-stage SMR developers is important. Existing facilities have operating histories and established regulatory frameworks. SMR developers may offer substantial long-term potential but face technology validation, licensing, construction, financing, and commercialization risk.

Natural gas generation and pipelines

Natural gas is positioned to provide much of the near-term flexibility required as electricity demand rises. Gas-fired plants can generally be developed faster than large nuclear projects and can operate as baseload, intermediate, or peaking resources depending on design and market conditions.

The opportunity extends beyond power plants to:

• Natural gas pipelines
• Storage facilities
• Gathering and processing infrastructure
• Liquefied natural gas infrastructure
• Gas turbines and related equipment

Representative examples include Williams, Kinder Morgan, Energy Transfer, Vistra, and GE Vernova. These companies have different business models and should not be viewed as equivalent investments.

Natural gas also introduces direct commodity and policy exposure. Fuel prices, pipeline constraints, emissions regulation, permitting, and changing market rules can materially influence returns. Gas may be essential to near-term reliability while still facing long-term competition from nuclear, renewables, storage, and demand-response technologies.

Regulated utilities

Regulated electric utilities may benefit from rate-base growth as they invest in generation, transmission, distribution, substations, and other system upgrades. If regulators approve prudent capital expenditures, utilities can earn an authorized return over time.

Examples include American Electric Power, Duke Energy, Southern Company, Dominion Energy, and NextEra Energy. Broad utility ETFs such as the Utilities Select Sector SPDR Fund (XLU) and Vanguard Utilities ETF (VPU) offer diversified exposure, although their holdings may include companies with limited direct AI-related power exposure.

Utility investing requires close attention to jurisdiction. A company operating in a state with supportive regulators and strong data-center demand may have a different outlook from one facing political resistance to rate increases or new infrastructure.

Transmission, grid equipment, and interconnection

Transmission and grid equipment may be among the most direct ways to express the power bottleneck. New data centers require transformers, switchgear, substations, power-management systems, high-voltage equipment, and transmission capacity.

Representative companies include Eaton, Hubbell, Quanta Services, and GE Vernova. The U.S. Energy Information Administration’s discussion of natural-gas pricing hubs illustrates how regional infrastructure constraints can affect energy markets. Similar constraints exist in electricity, where location and network capacity often determine whether power can be delivered economically.

This theme overlaps with the next article in the series, which will examine AI Physical Infrastructure, including data centers, cooling, construction, and related equipment. The distinction is useful: the power layer concerns generation, fuel, transmission, and grid access; the infrastructure layer concerns the facilities that consume and distribute that power.

Principal Risks Require Careful Underwriting

Power investments are not a one-directional AI trade. Key risks include:

• Regulatory and rate-case risk: Utilities must obtain approval to recover capital expenditures from customers. Regulators may reject projects, limit allowed returns, or require cost-sharing.
• Nuclear construction risk: New nuclear projects can experience delays, cost overruns, financing pressure, and licensing complications.
• Natural gas commodity exposure: Gas prices and regional pipeline constraints affect generation economics and margins.
• Interest-rate sensitivity: Regulated utilities and infrastructure companies often carry substantial debt and may trade partly as income-oriented equities. Higher interest rates can raise financing costs and reduce valuation support.
• Interconnection delays: A proposed data center may not receive power when expected because transmission studies, permitting, or equipment procurement take longer than anticipated.
• Demand-forecast risk: Efficiency improvements, changes in AI architectures, economic weakness, or slower data-center deployment could reduce expected load growth.
• Customer concentration: Projects built around one hyperscaler or a small number of counterparties face credit, renegotiation, and deployment risks.
• Overlapping exposure: An investor may own the same AI-power theme through a utility, a grid-equipment company, a nuclear ETF, and a broad technology fund without recognizing the concentration.

Is This a Near-Term Trade or a Long-Term Secular Theme?

It is both, but the investment case must separate the time horizons.

The near-term opportunity is an infrastructure buildout. Utilities, generators, pipeline operators, equipment manufacturers, and engineering firms may benefit from immediate spending to connect new loads and reinforce constrained systems.

The long-term opportunity is a secular increase in electricity demand. AI adoption is likely to continue expanding across enterprise software, industrial automation, scientific research, digital services, and consumer applications. Even if individual forecasts prove too aggressive, the underlying demand could persist for decades.

That does not mean every power-related company will outperform. Capital-intensive industries can destroy value when projects are overbuilt, financing costs rise, or regulators do not permit adequate returns. Investors must evaluate balance sheets, contracted revenue, regulatory jurisdictions, construction schedules, fuel exposure, valuation, and the credibility of management’s growth assumptions.

Portfolio Relevance: Focus on Exposure, Not Labels

Power is a foundational AI investment theme because every downstream layer depends on electricity. However, it is also a reminder that AI exposure should be analyzed across the full portfolio.

An investor may own a semiconductor manufacturer, a data-center operator, a utility, a grid-equipment supplier, and a broad technology index fund. Those positions may appear diversified by company name while remaining economically linked to the same AI-capital-spending cycle. This is the danger of double-counting AI exposure.

A disciplined review should identify:

• Direct AI holdings
• Indirect beneficiaries of data-center construction
• Utility and infrastructure exposure
• Commodity and interest-rate sensitivity
• Geographic and regulatory concentration
• The portfolio’s overall dependence on continued AI capital spending

For households with substantial assets, this analysis should connect to the broader financial plan. Concentrated equity exposure can affect liquidity, tax management, charitable giving, retirement distributions, and estate-planning decisions. The objective is not to predict which power technology will dominate. It is to understand how the exposure fits within the household’s risk capacity and long-term priorities.

Power may be the bottleneck for AI growth, but it is not a risk-free investment theme. Generation, fuel, transmission, and regulation each create separate return drivers. Investors who distinguish those drivers can pursue greater clarity, control, and portfolio efficiency.

Continue to the next layer: AI Physical Infrastructure

The examples of companies and exchange-traded funds in this article are provided for educational and illustrative purposes only. They are not recommendations, solicitations, or offers to buy or sell securities. Past performance does not guarantee future results. All investments involve risk, including possible loss of principal. Energy and utility investments may be affected by regulation, commodity prices, interest rates, construction costs, environmental policy, technological change, and market volatility. Forecasts and estimates are subject to change and may not be realized. Investors should consider their objectives, risk tolerance, tax situation, time horizon, and overall portfolio before making investment decisions. Please consult with a qualified financial professional regarding your individual circumstances. InSight Financial Planners is a Registered Investment Adviser. Advisory services are offered only where properly licensed or exempt from licensure.