AIEI Special Report #1
The Missing Category of AI Investing
Why AI Energy & Infrastructure Deserves Its Own Category
Research Brief • June 2026
Executive Summary
Artificial intelligence has become one of the defining economic and investment themes of the modern era.
Investors seeking AI exposure have largely concentrated attention on a familiar set of companies: semiconductor manufacturers, hyperscale cloud providers, foundation model developers, and software platforms. This focus has been logical. These organizations sit at the center of intelligence creation.
Yet a growing body of evidence suggests that another layer of the AI ecosystem is becoming increasingly important.
Every AI workload ultimately depends upon physical infrastructure.
Models require power.
Data centers require cooling.
Electricity must be generated, transmitted, conditioned, and delivered.
Equipment must be manufactured and deployed.
Infrastructure projects must navigate workforce constraints, permitting processes, and public acceptance.
As artificial intelligence scales, these systems become increasingly important alongside the systems creating intelligence itself.
This report argues that the energy and infrastructure layer supporting AI deployment represents a distinct category worthy of separate study.
The objective is not to construct an index, recommend securities, or propose an investment product.
The objective is to define the physical systems that increasingly determine how quickly intelligence can be deployed.
Introduction
The AI Waves framework was developed around a simple observation:
Artificial intelligence increasingly behaves less like a software cycle and more like a long-duration infrastructure buildout.
The first phase of the AI cycle focused on intelligence creation. Investors sought exposure to semiconductors, cloud infrastructure, foundation models, and the companies building increasingly capable systems.
These remain essential components of the AI ecosystem.
Yet as intelligence scales, new constraints emerge.
Power availability.
Grid capacity.
Electrical equipment.
Cooling systems.
Workforce availability.
Public acceptance.
The question is no longer simply:
Which companies are building intelligence?
An equally important question is:
Which companies are enabling intelligence deployment?
This report argues that the infrastructure supporting AI deployment represents a distinct category worthy of separate study.
AI Waves Framework
Wave 1 focuses on Compute & Intelligence Creation.
Wave 2 focuses on Energy & Infrastructure.
Future waves examine additional layers of intelligence deployment, integration, and physical-world adoption.
This paper focuses exclusively on Wave 2.
Scope of This Paper
This report is a category-definition paper.
Its purpose is to evaluate whether the infrastructure supporting artificial intelligence deployment represents a coherent category worthy of separate analysis.
It evaluates AI deployment constraints primarily through a U.S. lens, reflecting available infrastructure, regulatory, and capital-market data.
The report does not attempt to:
Construct an index
Recommend securities
Evaluate valuation levels
Assess expected returns
Establish portfolio allocations
Define methodology rules
Those questions require separate empirical and methodological work.
The objective here is narrower:
To determine whether AI Energy & Infrastructure represents a coherent category within the broader AI ecosystem.
AI Investing Has Become Concentrated
The market has become remarkably efficient at identifying companies associated with artificial intelligence.
Most AI-focused portfolios today contain substantial exposure to a familiar group of businesses:
Nvidia
Microsoft
Alphabet
Amazon
Meta
While individual products vary, many AI investment strategies ultimately concentrate around the same underlying themes:
Compute
Cloud Infrastructure
Software Applications
Model Development
This concentration has produced significant investment success.
It has also concentrated attention on the creation of intelligence.
Far less attention has been devoted to the systems required to deploy intelligence at scale.
History offers useful parallels.
Railroads required steel.
Automobiles required roads.
Electrification required transmission networks.
In each case, the visible technology depended upon a supporting infrastructure layer.
Artificial intelligence may prove no different.
As deployment scales, supporting infrastructure becomes increasingly important to outcomes across the broader ecosystem.
This raises a critical question:
If the market has become highly effective at identifying companies that build intelligence, has it become equally effective at identifying the systems that enable intelligence deployment?
The Missing Category
Artificial intelligence does not operate independently.
Every model, application, and inference request ultimately depends upon physical systems.
This dependency is easy to overlook because software remains the most visible layer of the AI stack.
Users interact with applications.
Developers interact with APIs.
Enterprises interact with software.
Yet beneath every interaction sits a physical infrastructure layer.
Electricity must be generated.
Power must be transmitted.
Equipment must condition and distribute energy.
Cooling systems must manage thermal loads.
Without these systems, intelligence cannot operate regardless of model quality or computational capability.
This report defines that collection of systems as:
AI Energy & Infrastructure
AI Energy & Infrastructure consists of companies whose products or services help relieve binding constraints on the large-scale deployment of intelligence.
This distinction matters.
Many industries may benefit from AI growth.
Far fewer directly address the constraints that determine how quickly intelligence can be deployed.
Wave 2 is defined not by exposure to AI demand, but by participation in solving deployment bottlenecks.
In other words, benefiting from AI growth is not sufficient. A company must help relieve a constraint that limits the deployment of intelligence.
Defining Wave 2
Within the AI Waves framework, Wave 2 represents the Energy & Infrastructure Layer.
Wave 2 consists of the systems that generate, transport, condition, store, and manage the resources required to deploy intelligence at scale.
The Wave 2 deployment framework is currently organized around six primary constraints:
Power Availability
Demand Growth
Equipment Capacity
Grid Capacity
Workforce Capacity
Public Acceptance
Wave 2 is not a traditional industry classification.
It is a deployment category.
Wave 2 companies are not grouped because they belong to the same industry.
They are grouped because they help relieve the same deployment constraints.
A utility, an electrical equipment manufacturer, a cooling provider, and an infrastructure contractor may appear unrelated through a traditional sector lens.
Viewed through the lens of AI deployment, however, they perform complementary functions within the same system.
The category is therefore defined by function rather than industry classification.
Its purpose is to identify the organizations helping relieve the deployment constraints that limit how quickly intelligence can scale.
Boundary Testing
A category is defined as much by its exclusions as its inclusions.
The purpose of Wave 2 is not to identify every company that may benefit from artificial intelligence.
The purpose is to identify companies that help relieve deployment constraints.
Example: Steel Producers
Steel demand may increase as AI infrastructure expands.
However, steel itself is not currently considered a primary deployment bottleneck.
Steel producers therefore do not automatically qualify as Wave 2 participants.
Example: Data Center Real Estate
Real estate may benefit from AI demand growth.
However, ownership of physical property alone does not necessarily relieve a deployment constraint.
Eligibility depends on whether the business directly addresses a critical infrastructure bottleneck.
Example: Semiconductor Manufacturers
Semiconductors remain essential to artificial intelligence.
However, they primarily contribute to intelligence creation rather than infrastructure deployment.
They are therefore generally classified within Wave 1 rather than Wave 2.
Principle
Benefiting from AI growth is not sufficient.
Relieving a deployment constraint is the defining characteristic of Wave 2.
The Constraint Stack Is Already Visible
The purpose of Wave 2 is not to predict future bottlenecks.
It is to identify bottlenecks that are already emerging.
The evidence increasingly suggests that AI deployment is constrained not by a single factor, but by a stack of interdependent physical systems.
These constraints do not operate independently. Delays in power generation can slow data-center construction, which can increase demand for electrical equipment, which can intensify labor requirements and permitting challenges.
The deployment stack behaves as a system rather than a collection of isolated bottlenecks.
The Deployment Constraint Stack
AI infrastructure does not depend upon a single bottleneck.
It depends upon interdependent constraints.
AI Demand
↓
Power Availability
↓
Grid Connection
↓
Electrical Equipment
↓
Construction & Deployment
↓
Operational Capacity
A delay at any stage can affect the stages that follow.
The deployment challenge therefore behaves more like a system than a collection of isolated bottlenecks.
Constraint 1: Power Availability
According to the Electric Power Research Institute (EPRI), data centers currently account for approximately 4–5% of U.S. electricity consumption. By 2030, that figure could rise to between 9% and 17% under current deployment scenarios.[1]
The implication is straightforward.
The limiting factor is increasingly not whether large technology companies can afford power.
The limiting factor is whether power can be secured quickly enough.
Constraint 2: Demand Growth
The International Energy Agency estimates that the United States could add more than 420 TWh of electricity demand through 2030, with data centers expected to account for roughly half of that increase.[2]
Data centers are no longer simply participating in electricity demand growth.
They are becoming one of its primary drivers.
Constraint 3: Equipment & Industrial Capacity
Power generation alone does not solve deployment constraints.
Electricity must still be transmitted, conditioned, distributed, and managed.
Wood Mackenzie projects that the U.S. data-center electrical equipment market could expand from approximately $20 billion to roughly $65 billion by 2030.[3]
In accelerated deployment scenarios, data centers could account for as much as 40% of the U.S. electrical equipment market.[3]
Constraint 4: Grid Capacity
Electricity must not only be generated.
It must also be delivered.
Grid infrastructure therefore represents a distinct constraint layer.
ERCOT has established dedicated procedures for large-load customers seeking 75 MW or greater of interconnected capacity.[4]
The existence of a specialized process reflects a broader reality.
Large AI deployments increasingly resemble infrastructure projects rather than traditional commercial loads.
Transmission capacity, interconnection timelines, substation development, and reliability considerations are becoming increasingly important.
Supporting Constraint: Deployment Capacity
Infrastructure ultimately depends on people.
Electricians, linemen, engineers, construction teams, and permitting professionals all play a role in bringing new capacity online.
Workforce availability may become increasingly important as AI infrastructure projects compete with broader electrification, industrial expansion, and grid modernization initiatives.
While labor constraints are difficult to quantify consistently across regions, they remain an important factor influencing deployment timelines.
Constraint 5: Public Acceptance & Regulation
Infrastructure projects require permits, approvals, and public support.
In a 2026 Reuters/Ipsos survey, only 33% of respondents supported rapid AI data-center expansion, while 77% expressed concern that additional data centers could increase electricity costs.[5]
As data-center development expands into new regions, infrastructure deployment increasingly becomes a political and regulatory issue rather than a purely technical one.
Several states and local jurisdictions have begun debating restrictions, permitting delays, or additional review processes for large-scale data-center projects, illustrating how deployment constraints can emerge through political as well as technical channels.
Key Observation
The defining challenge of the next phase of AI may not be creating intelligence.
It may be deploying intelligence.
Power availability.
Demand growth.
Equipment capacity.
Grid capacity.
Workforce Capacity.
Public acceptance.
Each represents a distinct bottleneck.
The companies grouped within Wave 2 do not belong to the same industry.
What links them is their role within the same deployment system.
Illustrative Boundaries
The companies listed below are illustrative rather than exhaustive.
Representative participants include:
Constellation Energy
Vistra
NextEra Energy
Eaton
GE Vernova
Schneider Electric
Vertiv
Trane Technologies
Quanta Services
These companies operate in different industries and sectors.
What connects them is their functional role within the deployment of intelligence.
Conclusion
Artificial intelligence is often discussed as a software story.
Increasingly, it is also becoming an infrastructure story.
The systems required to generate, deliver, condition, cool, and support electricity are becoming more important as intelligence scales.
This report proposes a different lens.
Rather than grouping companies by sector, it groups them by function within the deployment of intelligence.
The evidence presented here suggests that power availability, grid capacity, equipment capacity, deployment capacity, and regulatory considerations are becoming increasingly relevant to the pace of AI deployment.
AI Energy & Infrastructure is not defined by a shared industry classification.
It is defined by a shared role in relieving the constraints that increasingly shape the deployment of intelligence.
If deployment becomes as important as creation, then understanding those systems becomes increasingly important as well.
That is the purpose of AI Energy & Infrastructure.
Not to replace existing categories.
But to provide a framework for understanding a layer of the AI ecosystem that is becoming increasingly difficult to ignore.
References
[1] Electric Power Research Institute (EPRI). Data Centers Could Consume Up to 17% of U.S. Electricity by 2030. February 2026.
[2] International Energy Agency (IEA). Electricity 2026: Demand Outlook. 2026.
[3] Wood Mackenzie. Data Center Demand Drives U.S. Electrical Equipment Market to $65 Billion, Reshaping Industry Dynamics. 2026.
[4] Electric Reliability Council of Texas (ERCOT). Large Load Integration Framework. Accessed June 2026.
[5] Reuters/Ipsos. Americans Wary of AI-Driven Data Center Boom. June 2026.
Related Research
Coming Soon
AI Energy Index Methodology