Here’s how scarce electricity could hamper the AI investing boom

We think the building of new power generators offers attractive return potential at a time when $600 billion in transmission and distribution spending is projected through 2030.⁴ This cycle may be long-lived, as the power sector has experienced about $3 trillion in underinvestment since 1999.⁵ The high, growing and unmet demand for power can’t be fulfilled quickly.

We view the cycle favorably, particularly select power generators, “picks and shovels” (engineering and construction) firms, electrical equipment manufacturers and others sitting close to the deployment bottleneck.

Here, we examine the structural forces driving the power build-out and the parts of the ecosystem most likely to benefit. We see the most compelling potential opportunity in decades in utilities—but only if investors exercise selectivity. Outcomes will vary widely by local power-market constraints and regulations, which could extend the demand shock through the end of this decade. We consider the possible upside forces and potential risk signals that may help us identify emerging leaders across a handful of industrial sectors.

Hurdles: Practical, political and technical

The electrical grid, a physical system—generators, interconnection, the grid, equipment manufacture—is growing capacity at a linear pace. By contrast, AI-related computing needs are scaling exponentially.

Hyperscalers and power project developers are looking to build in regions with available capacity, faster interconnection and supportive policy. They face many hurdles: A lack of skilled labor, complex regulations, long production times for critical equipment like turbines and large transformers, and growing popular and political opposition to data centers and the rising electricity costs to consumers that many have triggered.

Satisfying AI’s demand for power will require a mix of fuel sources, including nuclear and renewables, each with pros and cons. Data center operators and developers can get online on the grid faster, and ease some opposition, by building and paying for their own power plants, called dedicated or “behind-the-meter” supply,⁶ which can reduce how much of the bill lands on existing utility customers. That’s a trend we expect will grow. But even then, they must still negotiate permits, connection approvals, reliability rules set by regulators, and potentially a multi‑year wait.

Regulation and politics: Key variables informing the outlook

Regulation is now in focus. In tight power markets, regulatory rules are more likely to determine power generators’ outcomes than fuel costs or technology choices. The demand growth matters for portfolio decisions, but clarity around key regulations is a leading driver of investor outcomes.

As the power system moves from stable demand to a steep, step-function growth in demand, it’s not unusual to see an industry confront the constraints with transitional planning and turn to emergency procurement mechanisms. Regulators are stepping in, but there isn’t a clear path forward yet. As we’ll discuss, nonprofit PJM Interconnection, the largest U.S. regional grid operator running the power market from New Jersey to Illinois, is a real-time case study. 

Politics are another major constraint. Someone pays in conditions of scarcity, and in some districts where that has fallen to consumers, voters have become upset about affordability. Regulatory proceedings have recently been allocating the cost of new generation, among data centers, residential ratepayers and utility shareholders. At the time of this writing, the regulatory environment remains uncertain.

Still, this is potentially the most interesting opportunity for investors in the utilities and adjacent sectors in decades. But only, we think, with selectivity. We do not believe the impacts will be equally distributed. Here’s how we understand the unique local environments where power is generated, including the complex and extremely regulated ways electricity comes to market. The constraints and bottlenecks could, we think, prolong the demand shock and the investing opportunity through the end of the decade.

To be sure, demand growth could undershoot current projections if chip energy efficiency, the architecture of AI models or workload optimization prove able to reduce the power intensity of AI computing. But we don’t expect efficiency gains to be enough to offset leading AI companies’ power requirements, which are expected to double by 2027, according to IEA estimates.

For further background on the opportunity, see Michael Cembalest’s Eye on the Market energy paper.

Money can’t buy you electric power

A power project’s debut often depends on its place in a years-long queue, rather than on capital. Large customers must get in line for a utility or grid operator to run engineering checks, make local upgrades, approve plans and schedule work.⁷

This isn’t a new story. As early as 2024, the increased demand for power from AI data centers had begun to take center stage. We predicted the unsatisfied demand in earlier reports and reiterated that demand growth was a high conviction theme in both public and private markets. Now demand growth is at an over 20-year high and U.S. electrical infrastructure can’t keep up. After 20 flat years of industrial growth, the situation has changed.⁸

A year ago we called it “a supply crisis putting grid reliability and AI expansion at risk” and “a risky imbalance” between critically tight supply and demand in 70% of regional power markets. We said rolling blackouts and higher prices were likely. These have come to pass.

Case in point: The story of PJM

Price volatility, demand-supply imbalance, popular dissent: All three big themes are clearly visible at PJM. In July 2025, PJM’s “capacity auction” (a forward market that pays generators for being available in future peak seasons) hit a record $329.17 per megawatt per day—a rise of more than 1,000% over the last two years. PJM is responsible for satisfying nearly half of U.S. data center power demand because so many are located in its region. Forecasts suggest data centers could comprise about one-fifth of total demand in the region within a decade.⁹

Increasing power prices and affordability concerns have been top headlines this year. Governors in the affected states are pushing for price caps and cost controls; PJM is undertaking various reforms and federal involvement has risen. Meanwhile PJM has sped up its interconnection queue, adding new supply through a reform of its processes.

Investible opportunities in the power capex cycle

For investors, the power capex cycle points to the likely rising value of new power that will bring large customers flexibility and reliability. One opportunity is in what’s called “fast-start” generation (“fast power”), sector-wise mostly heavy industrial companies. Another area of interest: storage, for intermittent renewable power. We also like grid build-out and improvement opportunities, and related services including software.

Here’s a look at a handful of key areas of interest:

Equipment suppliers (industrials—electrification and power equipment)

The industrials and heavy equipment suppliers are crucial in the race to build fast power. As noted, equipment and grid hardware manufacturing lead times typically are multi-year. The cost of combined-cycle gas turbines is also rising. Transformer and other electrical equipment lead times range from 18 to 36 months. Original equipment manufacturers (OEMs) and suppliers of generation sets, electrical balance-of-plant, and controls are opportunities in this area.

Positive signals:

• Natural gas is foundational because it is dispatchable, scalable and cost-competitive for round-the-clock electricity loads. Gas generation could more than double by 2030. Backlogs for gas turbine orders are years long in many cases, pushing delivery timelines late into this decade.
• Grid equipment suppliers are likely to benefit as they supply large transformers, switchgear,¹⁰ substation build-outs, protection/control and high-voltage transmission components, all of which are critical components that are not available quickly in sufficient quantities.
• Grid upgrades are underway and should accelerate after transmission line growth trended down for a decade, markedly below Department of Energy (DOE) targets.

What we’re watching:

If hyperscaler capex slows, the demand signal feeding the power build-out could weaken before supply catches up, changing the balance of risk for late-cycle equipment and capacity additions.

Engineering, procurement and construction (EPC)

These companies design and build power infrastructure: generation plants, substations and behind-the-meter systems. Those with picks and shovels near bottlenecks are likely to benefit soonest.

Positive signals

• Our outlook is especially favorable for companies building gas generators. Orders and backlogs are at all-time highs, and as noted above, natural gas generation is expected to double by 2030. Combined-cycle gas turbines (CCGTs) and gas peakers will bring more than two-thirds of the incremental supply likely to be added to meet data centers’ power demand through 2035.¹¹ Natural gas turbine orders have soared with record backlogs of up to three to seven years for CCGTs.

What we’re watching

• Construction bottlenecks may occur as builders await critical components, as noted.
• Skilled labor is not available quickly in sufficient quantities.
• The risk of overbuilding gas generation is a concern; once the constraint eases, could there be too many (not too few) plants, which happened in the late-1990s?
• Some data centers will shift workloads geographically, but many areas’ state and local governments are pushing back with moratorium attempts and other restrictions.

Regulated utilities and independent power producers

In public equity and bond markets, utilities and power producers remain key beneficiaries of this capex cycle. Independent power producers (IPPs) are positioned to benefit from rising prices in constrained markets, where reliable electrical capacity becomes more valuable. Regulated utilities¹² have the potential to benefit from increased capital investment and an expanding rate base tied to the power infrastructure build-out, and the earnings outlook reflects this shift.

Positive signals:

• Major investing taking place in grid upgrades.
• Guaranteed power supply agreements for hyperscale data centers have evolved into complex, multi-decade commitments. While long commitments could deplete AI companies’ cash flow, they should be a positive for utilities.

What we’re watching

• Questions abound: Will regulators allow utilities to earn returns in a timely way, to recover costs? Will regulators permit a satisfactory return on equity for transmission/distribution upgrades? Will political pressure force costs to be absorbed by the companies (not so much by consumers)? Will there be unexpected or prolonged delays? Investors will watch these matters closely.
• Another risk, as noted, is how soon, and by how much, data centers will learn to do more with less power.
• We are wary of speculative power plant investment without contracted tenants.

Second-order providers adjacent to the demand shock

A range of companies enable data centers to acquire and run fast power. These can include software and other service firms tied to interconnection services, providers of grid-modernization software and energy services that shorten time-to-power. (This category also includes inputs into the aforementioned picks and shovels companies, such as makers of electrical equipment.)

Positive signals:

• Equipment shortages—particularly in transformers and electrical systems—have extended lead times. The higher costs, which reflect both demand pressures and supply chain constraints, boost revenues for equipment producers.
• The battery market is expected to see growth because batteries can be installed and start delivering power relatively quickly. Batteries are increasingly being paired with solar projects (batteries combined with solar provide a more reliable solution to manage renewable variability and manage peak demand).
• The cost of battery systems is declining, making such projects more economical.

What we’re watching:

• Because of supply chain pressures, input costs face some risk of rising.
• Tariffs could amplify this for inputs produced outside the United States.

Renewable energy and other fuel sources

No single fuel source can serve the full spectrum of needs—reliability, scalability, speed and cost—on its own. We think an all-of-the-above approach is mandatory. Renewables can be deployed quickly but have intermittency issues. Natural gas will be key, but gas turbines have long backlogs. Nuclear restarts are sparse and new builds can take a decade. The lives of coal-fired plants are being extended instead of retired, but the longer-term outlook for the group remains negative.

Positive signals:

• More wind, solar and other sustainable energy sources are likely to serve the escalating power demand, but they are intermittent fuels. If these additions are less dependable in times of peak demand, the grid will need even more capacity to be reliable—potentially increasing total capex required and extending the duration of the power build-out cycle.
• Storage, which converts intermittent generation from sun or wind into a more usable steady supply, is crucial infrastructure and has become a core area in this capex cycle, as we’ve noted. Global storage deployments are projected to experience about a 20% compound annual growth rate in the next five years. 
  
• Nuclear power matches the hyperscalers’ load profile. But new builds aren’t expected to come online until the second half of the decade, making nuclear a longer-term solution. While small modular reactors may improve scalability over time, they are unlikely to materially address supply gaps in the near term, given that the technology is still in the early stage of development.

What we’re watching:

We are wary of overreliance on renewables as a standalone solution—a balanced approach is needed. (For example, grid hardening for 24/7 reliability and spending on grid modernization across the broad ecosystem, including transmission, substations, gas generators, storage and grid services.

Infrastructure and real assets

In private market investing, value creation around the power capex cycle initially has been in infrastructure, and power has expanded to become the largest infrastructure sector globally. Infrastructure investments may offer comparatively stable, long-term cash flows (sometimes supported by contracted and/or inflation-linked revenue) and can provide portfolio diversification, though outcomes vary by asset, structure, and market conditions. The broad infrastructure opportunity includes investments in essential physical and digital assets that underpin economies. 

Positive signals:

• Accelerating power demand, coupled with a grid in which several infrastructure categories are older than their expected lifespans, has led to concerningly constricted supply.
• The power capex supercycle is supported by the trend of improving grid resilience as a matter of national security (not just to support AI computing).
• Transmission companies and providers of goods and services to the grid build-out are the theme’s cleanest expressions.
• Decades of underinvestment have created at least a $550 billion capex opportunity¹³ that utilities’ balance sheets cannot fund alone. We favor these investments for their long-duration, regulated cash flows.

Investment implications

The shifts we’ve described are reshaping earnings trajectories, company valuations, capital flows and power infrastructure’s potential role in portfolios. Yet a large percentage of some investor groups have 0% infrastructure exposure (for example, family offices).

Consider whether it is timely to stop thinking of power a steady-state allocation.

We note that we don’t consider it a thematic trade because we expect a large degree of dispersion among companies in the space—and dispersion in performance, within and across subsectors, is rising. Not every firm will be able to execute given the constraints.

Public market equities

Consider engineering and construction firms, electrical equipment manufacturers, transformer suppliers and cooling/HVAC providers that sit closer to the deployment bottleneck than many generators do. These are in what we call the HALO category—heavy assets, low obsolescence.

We also favor some publicly traded independent power producers (IPPs) and utilities. Utilities sector earnings growth expectations have accelerated into the high single- to low double-digit range. The IPPs’ growth is well above the regulated utilities but the group is much more volatile. Regulated utilities, however, are likely to account for around 70% of the sector’s growth over the next two years, highlighting that strength potentially won’t only come from the IPPs.

Valuations remain reasonable, relative to the broader market. Dividend distributions (approximately a 3% dividend yield) are above market averages, too, creating an even more attractive total return potential. The sector’s composition makes it attractive for investors looking for exposure to the AI theme outside tech, and those seeking downside resilience and the potential for steady returns. 

Private markets

Infrastructure is our preferred implementation in this power capex cycle. Core infrastructure provides direct exposure to long-duration build-out needs, with long-term, contractual inflation-resilient cash flows—an important feature at a time of higher and more volatile inflation.

Power now represents nearly 60% of the global infrastructure benchmark index. Yet allocations remain surprisingly low: 79% of our family office clients, and 86% of other Private Bank clients, have 0% infrastructure exposure.

We think it’s time to reconsider that allocation, as transmission and the grid are structurally supported after decades of underinvestment. Participating in this timely capex cycle depends, we believe, on selectivity.