Why Power, Cooling, Generators, and Energy Storage Are Becoming Critical AI Infrastructure

📰 iAtlas Daily #38 | August 2026

📰 What Happened?

The AI boom is creating an unexpected group of winners.

Not only NVIDIA, semiconductor manufacturers, and cloud companies.

Traditional industrial companies are increasingly benefiting as the rapid construction of AI data centers creates enormous demand for electricity, generators, power equipment, cooling systems, and energy storage.

Companies including Caterpillar, Cummins, Eaton, and Ford are expanding their exposure to the data-center market.

Caterpillar is investing approximately $725 million to expand generator manufacturing capacity, while Cummins expects its data-center-related business to potentially reach $9 billion by 2030. Eaton has also invested heavily to expand its data-center infrastructure portfolio.

This reveals an important shift in the AI investment cycle.

The AI race is increasingly becoming a race to build the physical infrastructure underneath computing.

💡 Why This Matters

AI Has a Physical Infrastructure Problem

AI is often described as a software revolution.

But every AI model ultimately runs on physical hardware.

That hardware requires:

GPUs -> Servers -> Data Centers -> Power Distribution -> Cooling -> Electricity Generation

As AI models become larger and computing systems more powerful, the infrastructure underneath them must scale as well.

And power is emerging as one of the industry’s biggest constraints.

JLL estimates the global data-center sector could add around 100 GW of capacity by 2030, requiring roughly $3 trillion of investment across infrastructure, real estate and related financing.

⚡ The Bottleneck Is Moving From Chips to Power

For much of the AI boom, the biggest bottleneck was advanced semiconductors.

GPUs were difficult to obtain.

HBM capacity was limited.

Advanced packaging became constrained.

Those challenges have not disappeared.

But another constraint is moving rapidly up the list:

electricity.

AI data centers can require hundreds of megawatts—and future campuses are increasingly being discussed at gigawatt scale.

Connecting facilities of that size to existing electricity grids can take years.

That is forcing developers to consider alternative strategies:

On-site generation
Gas turbines
Large generators
Renewable energy
Battery storage
Microgrids

The data-center race is therefore becoming a power infrastructure race.

🏭 Caterpillar Finds a New AI Market

Caterpillar provides one of the clearest examples.

The company is traditionally associated with construction equipment, mining machinery and heavy industrial engines.

But AI data centers require enormous amounts of reliable backup and on-site power.

Caterpillar is responding by investing $725 million to increase generator production and bringing larger-capacity products to the data-center market.

The connection may initially seem surprising.

What does a heavy-equipment company have to do with AI?

Quite a lot.

AI infrastructure cannot operate without reliable electricity.

That makes generators and power systems part of the AI hardware ecosystem.

⚙️ Cummins Sees Billions in Data-Center Demand

Cummins is seeing a similar opportunity.

The company manufactures engines, generators and power systems used across industrial applications.

It expects its data-center-related sales could reach approximately $9 billion by 2030 and is expanding generator production to meet demand.

This represents a broader industrial trend.

The beneficiaries of AI investment are expanding beyond semiconductor companies.

They increasingly include companies that manufacture:

Generators → Transformers → Switchgear → UPS → Cooling → Power Electronics

AI is pulling traditional industrial equipment into the technology supply chain.

🔌 Eaton Shows Why Power Distribution Matters

Generating electricity is only part of the challenge.

Power must also be safely and efficiently distributed through a data center.

That requires:

Transformers
Switchgear
UPS systems
Circuit protection
Power distribution equipment
Monitoring systems

Eaton has spent approximately $13 billion on acquisitions as it expands its position in data-center infrastructure, and data-center-related operations now represent a significant portion of its business.

The opportunity grows as rack power density increases.

AI servers consume significantly more power than traditional computing infrastructure.

That means the electrical architecture itself must evolve.

🔋 Batteries Enter the AI Infrastructure Chain

This is where yesterday’s Daily #37 becomes especially relevant.

Battery companies are also entering the AI infrastructure ecosystem.

Ford is redirecting excess EV battery manufacturing capacity toward stationary energy-storage systems for data centers, utilities and industrial customers.

Samsung SDI is similarly expanding its ESS strategy as EV demand growth becomes less predictable.

The connection is increasingly clear:

AI Data Center -> Large Electricity Demand -> Grid Constraints -> On-Site Power + ESS
-> New Battery Demand

The AI boom is therefore creating a new connection between the semiconductor and battery industries.

❄️ Cooling Is Becoming Critical

Power entering a data center eventually becomes heat.

And AI servers generate a lot of it.

Traditional data centers relied heavily on air cooling.

But increasingly dense AI racks are pushing infrastructure toward more advanced thermal-management systems.

These include:

Direct-to-chip liquid cooling
Coolant distribution units
Heat exchangers
Chillers
Pumps
Advanced thermal controls

As rack density increases, cooling becomes not simply a facility requirement but part of computing performance.

A GPU that cannot be cooled efficiently cannot operate at maximum performance.

🔌 NVIDIA Is Changing the Power Architecture

The scale of the problem can also be seen in NVIDIA’s infrastructure roadmap.

On August 11, NVIDIA detailed its 800 VDC power architecture for next-generation AI factories.

The goal is to reduce power-conversion complexity and improve efficiency as AI rack density continues increasing.

Traditional data centers convert electricity multiple times before it reaches computing hardware.

At very high power levels, every conversion creates additional losses and complexity.

Higher-voltage DC distribution can simplify that path.

This demonstrates how deeply AI is beginning to influence electrical engineering itself.

🌍 The Industrial Supply Chain Around AI

The AI infrastructure ecosystem is becoming remarkably broad.

💻 Semiconductor

GPUs, CPUs, networking chips and HBM provide computing capability.

⚡ Electrical Equipment

Transformers, switchgear, UPS and power-distribution systems deliver electricity.

🏭 Power Generation

Generators and turbines provide backup or on-site electricity.

❄️ Cooling

Liquid cooling and thermal-management systems remove enormous quantities of heat.

🔋 Energy Storage

ESS provides backup power, load management and renewable-energy integration.

🧪 Materials

Copper, steel, aluminum and specialty materials are required throughout the infrastructure.

The AI boom is therefore moving far beyond the technology sector.

Korean Manufacturers Are Seeing the Same Opportunity

This shift is not limited to the United States.

Korean industrial companies are also repositioning around AI infrastructure.

Recent reporting shows Korean steel, petrochemical, electrical-equipment and engineering companies targeting opportunities in areas including specialty steel, ESS, power equipment and data-center infrastructure.

This could become particularly important for Korea.

The country already has strong industrial capabilities across:

Semiconductors + Batteries + Electrical Equipment + Heavy Industry + Engineering

AI infrastructure potentially connects all five.

📈 From AI Chips to AI Factories

The AI investment cycle can now be viewed in several layers.

Layer 1 — AI Models

Software and algorithms.

Layer 2 — Compute

GPU, HBM and networking.

Layer 3 — Data Centers

Servers and computing infrastructure.

Layer 4 — Physical Infrastructure

Power, cooling, buildings and energy storage.

Layer 5 — Industrial Supply Chain

Generators, transformers, batteries, steel, copper and engineering.

Each layer creates demand for the one underneath it.

This is why the economic impact of AI is becoming much broader than the semiconductor industry.

👀 What to Watch

Several indicators deserve attention:

AI data-center electricity demand
Generator and turbine orders
Transformer availability
Liquid-cooling adoption
Data-center ESS deployments
On-site power generation
800 VDC infrastructure adoption
Grid connection delays
Data-center construction spending

But perhaps the most important metric will be:

Time to power.

Having GPUs available means little if a new data center cannot secure enough electricity to operate them.

🎯 Atlas Insight

The first phase of the AI boom rewarded companies that could create computing power.

The next phase may increasingly reward companies that can deliver physical power.

That distinction matters.

The AI supply chain now looks increasingly like:

AI Models -> GPU + HBM -> Servers -> Data Centers -> Power + Cooling -> Generators + ESS + Grid Infrastructure -> Industrial Materials & Equipment

This explains why companies traditionally considered part of the industrial economy are suddenly becoming part of the AI story.

Caterpillar does not manufacture GPUs.

Eaton does not train AI models.

Ford is not a cloud-computing company.

Yet all three can benefit from AI infrastructure investment.

AI is no longer creating only a semiconductor boom. It is beginning to create an industrial equipment boom.

And that may become one of the most important industrial trends of the second half of this decade.

📚 Related Articles

📰 iAtlas Daily #33 | NVIDIA and Wall Street Target $500 Billion for AI Infrastructure

📰 iAtlas Daily #36 | Applied Materials Expands for the AI Chip Boom

📰 iAtlas Daily #37 | Samsung SDI Takes Full Control of GM Battery Plant

📊 iAtlas Weekly #6 | Scale Is Becoming the New Industrial Advantage

Sources

About iAtlas

iAtlas is an independent publication covering batteries, semiconductors, OLED, advanced materials, AI, and global industrial trends.

We transform complex industrial developments into clear, reliable, and easy-to-understand insights.

Whether you’re following today’s industry news or building long-term expertise, iAtlas helps you understand not only what happened, but why it matters.

Technology creates change.
Insight creates opportunity.
— iAtlas

AI Data Centers Are Creating a New Industrial Boom