AI Waves™ Inaugural Edition (v1.1)

Tracking How AI Constraints Migrate Into Physical Infrastructure Systems Over Time

Research Brief • June 2026

Abstract

AI Waves™ is a research framework examining how AI constraints migrate from software and compute into physical infrastructure systems.

This inaugural edition explores the progression from compute bottlenecks to energy, transmission, cooling, coordination, labor, and physical-world deployment.

The central thesis: AI increasingly behaves less like a software cycle and more like a long-duration infrastructure buildout.

Executive Summary

The dominant assumption in technology cycles is that AI scaling is primarily a software and compute challenge.

That assumption is increasingly being revised.

As AI systems grow in scale and deployment, the constraints shaping their trajectory are migrating outward—from model architectures and accelerator availability into the physical systems required to power, cool, and transmit AI at industrial scale.

AI Waves™ is a framework for tracking how those dependencies evolve over time.

Core Observation

Compute moves in quarters. Infrastructure moves in years.

This mismatch may become one of the defining characteristics of the AI era.

The Constraint Migration Thesis

AI bottlenecks do not disappear when solved.

They migrate.

As compute capacity expands through capital investment, pressure shifts downstream into the physical systems required to sustain deployment at scale.

Early constraints centered on:

• Semiconductor supply

• Accelerator access

• Data center capacity

As those improve, pressure increasingly moves into:

• Electrical infrastructure

• Substations

• Transmission systems

• Thermal management

• Large-scale operational coordination

Each solved bottleneck exposes the next.

The Five Waves Framework

Wave 1 — Compute Buildout

Status: Active + Observable

The first wave of AI expansion was defined by compute.

Hyperscaler investment, accelerator deployment, and training infrastructure formed the origin layer of the current cycle.

Wave 2 — Energy & Physical Infrastructure

Status: Active + Observable

As compute scales, electricity and physical infrastructure increasingly become the binding constraint.

What begins as a software challenge gradually becomes a power, transmission, cooling, and construction challenge.

Key dependencies include:

• Electricity generation

• Substations

• Transmission

• Cooling infrastructure

Wave 3 — Operational Ecosystems

Status: Emerging + Monitored

As infrastructure scales, coordination becomes increasingly important.

The challenge shifts from building infrastructure to orchestrating it efficiently across complex systems.

Areas being monitored:

Established

• Flexible load balancing

• Industrial coordination software

Adapting

• Energy-aware orchestration

Emerging

• Digital twin infrastructure

• AI-native orchestration

Wave 4 — Sector Intelligence Systems

Status: Long-Duration Domain

As intelligence becomes embedded into real-world systems, AI increasingly migrates into sector-specific operating environments.

Areas monitored:

• Healthcare

• Logistics

• Industrial systems

• Defense

Wave 5 — Civilization Infrastructure Coordination

Status: Emerging Signals

Physical AI, robotics, autonomous systems, and machine-operated infrastructure are beginning to move from conceptual discussion into observable deployment.

The timeline may be closer than many assume.

Areas monitored:

• Energy systems

• Compute systems

• Logistics networks

• Autonomous coordination

• Physical AI and robotics

Constraint Migration in Practice: Northern Virginia & PJM

Northern Virginia provides a real-world example of constraint migration.

The region hosts the largest concentration of data center infrastructure in the world, accounting for roughly 25–30% of U.S. hyperscale capacity.

The region illustrates how bottlenecks evolve over time.

Phase 1 — Compute Demand Concentrates

Hyperscaler investment accelerated rapidly between 2021 and 2024.

Phase 2 — Power Systems Become the Constraint

Load growth began outpacing substation and transmission capacity.

Phase 3 — Transmission Limits Emerge

Large-load interconnection queues expanded while infrastructure timelines stretched into multi-year horizons.

Northern Virginia is likely not an outlier.

It may be an early indicator of patterns beginning to emerge in other regions, including Texas.

Emerging Constraint: Skilled Labor

AI infrastructure deployment depends on:

• Electricians

• Linemen

• HVAC technicians

• Construction crews

• Welders

• Engineers

• Project managers

Even if capital and energy become available, workforce capacity may emerge as an independent bottleneck.

This remains an underappreciated area of risk.

A Synchronization Problem

A fundamental mismatch exists between software deployment timelines and infrastructure deployment timelines.

Compute Moves Quickly

• Months

• Global supply chains

• Modular deployment

Infrastructure Moves Slowly

• Multi-year permitting

• Regional constraints

• Sequential construction requirements

Compute moves in quarters. Infrastructure moves in years.

This may become one of the defining tensions of the AI era.

Risks, Timing & Counterforces

The framework does not assume the thesis is inevitable.

Several counterforces may alter the trajectory.

Efficiency Gains vs. Jevons Dynamics

Efficiency improvements may reduce resource intensity while simultaneously increase overall demand.

Historically, improved efficiency has often expanded consumption rather than reducing it.

Regional Variability

Infrastructure constraints differ significantly by geography.

Some regions may adapt rapidly while others face prolonged bottlenecks.

Capital Mobilization

Capital is responding aggressively.

Deployment timelines remain uncertain.

Deployment Timeline Risk

Permitting, labor, transformers, cooling equipment, and transmission all introduce timing risk.

Execution may ultimately matter more than capital availability.

Closing Thesis

AI may increasingly behave less like a software cycle and more like a long-duration infrastructure buildout.

The central observation of AI Waves™ is not that AI will transform civilization.

The observation is narrower and more operational.

AI deployment at scale increasingly exhibits the characteristics of infrastructure expansion.

It depends on physical systems.

It operates on construction timelines.

It creates regional bottlenecks.

It accumulates long-duration capital dependencies.

Sources

Primary and secondary source categories referenced in this research brief include:

• EPRI

• FERC Order 2023

• PJM Interconnection

• EIA

• Lawrence Berkeley National Laboratory

• Rocky Mountain Institute

• McKinsey Global Institute

• S&P Global

• Reuters

• Utility and Grid Research

Related Research

Coming Soon

• AI Energy Index Methodology

• AIEI Special Report #1

• AI Waves Vol. 2

Download

Download the full PDF version of this research brief.