Why Wall Street is Pivoting From AI Apps to AI Hard Assets

Why is Wall Street pivoting from AI apps to AI hard assets? Because artificial intelligence increasingly depends on scarce physical infrastructure: GPUs, semiconductors, data centers, electricity, cooling systems, grid connections, fiber networks and land. As spending on AI infrastructure accelerates, investors are looking beyond the applications using artificial intelligence to the companies supplying the physical resources every AI developer needs.

This doesn’t mean Wall Street has abandoned AI software. Instead, the investment thesis is broadening from:

“Which AI app will win?”

to:

“Who supplies the infrastructure every AI winner must use?”

The numbers explain the enthusiasm. The International Energy Agency (IEA) says capital expenditure by five major technology companies exceeded $400 billion in 2025 and is expected to increase another 75% in 2026, driven by data-center investment.

Microsoft alone says it expects roughly $175 billion of calendar-year 2026 capital expenditure after accounting for a lease-classification change.

The AI boom is therefore increasingly becoming a physical infrastructure boom.

AI Apps vs AI Hard Assets

1. AI Needs Enormous Physical Infrastructure

Artificial intelligence might appear entirely digital. It isn’t. Every prompt ultimately runs through physical equipment located somewhere in the world.

That requires:

Semiconductors → Servers → Data centers → Electricity → Cooling → Networks

The IEA estimates that traditional data centers generally use 10–25 megawatts of electricity, while hyperscale AI facilities can exceed 100 MW.

A 100-MW facility can consume electricity equivalent to roughly 100,000 households. Some of the largest facilities under construction could consume as much electricity as approximately 2 million households.

This changes the investment question.

The bottleneck may not be creating another AI application. It may be finding enough electricity and computing infrastructure to run thousands of them.

2. Data Centers Are Becoming the Factories of the AI Economy

Think of data centers as AI factories. Traditional factories transform:

Raw materials → Manufactured goods

AI data centers transform:

Electricity + Compute + Data → Intelligence

That transformation requires enormous capital expenditure.

The IEA estimates global investment in data centers had already nearly doubled since 2022 to approximately $500 billion in 2024. Investment has continued accelerating.

Microsoft reported $41 billion in quarterly capital expenditure in its fiscal 2026 fourth quarter. Roughly two-thirds went toward shorter-lived assets, primarily CPUs and GPUs, while part of the remaining spending supported longer-lived infrastructure such as data-center sites.

That is why physical AI infrastructure has attracted so much investor attention.

3. Electricity Is Becoming an AI Asset

AI cannot scale without electricity. This may be one of the most important investment themes of the next decade. According to the IEA, global data-center electricity consumption reached approximately 485 terawatt-hours in 2025.

By 2030, it projects consumption of approximately:

950 TWh

That’s roughly double today’s level and approximately 3% of global electricity demand. More strikingly, electricity consumption from AI-focused data centers grew approximately 50% in 2025, according to the IEA.

The investment chain therefore expands:

AI demand → More GPUs → More data centers → More electricity → More generation → More transmission → More grid equipment

An investor betting on AI doesn’t necessarily need to predict which chatbot dominates. They can potentially examine the infrastructure supplying the entire ecosystem.

4. GPUs and Semiconductors Are the Picks and Shovels

During a gold rush, selling picks and shovels can sometimes be more predictable than searching for gold. AI semiconductors represent a modern version of that idea.

Read Also: The Geopolitics of AI Infrastructure: National Sovereign Clouds

Thousands of AI applications can fail. But collectively they still require:

• GPUs.
• CPUs.
• Memory.
• Networking equipment.
• Storage.
• Semiconductor manufacturing.

Microsoft says roughly two-thirds of its $41 billion quarterly capex in its latest reported quarter went toward shorter-lived assets, predominantly CPUs and GPUs. The infrastructure thesis is therefore based partly on shared demand.

Instead of asking:

Which AI application wins?

investors can ask:

What does virtually every serious AI company need?

Compute is near the top of the list.

5. The Electrical Grid Could Become an AI Bottleneck

Building a data center isn’t enough. You must connect it to electricity.

That means AI investment increasingly intersects with:

Utilities + Transformers + Transmission + Switchgear + Substations + Backup generation + Energy storage

Data-center demand is already changing electricity consumption. The IEA reports that rapidly expanding data centers contributed roughly half of the increase in U.S. electricity consumption during 2025.

Through 2030, the agency expects data centers to account for almost half of U.S. electricity-demand growth. That makes electricity infrastructure an indirect AI investment theme.

The AI economy cannot grow faster than the infrastructure supplying its power indefinitely.

6. Natural Gas, Renewables and Nuclear Enter the AI Story

Where will all that electricity come from? There is unlikely to be one answer. The IEA expects renewables and natural gas to lead in meeting additional global data-center electricity demand, while other energy sources also contribute.

This expands the AI infrastructure theme beyond technology. Potential beneficiaries of rising power requirements can include industries connected to:

Natural gas + Solar + Wind + Nuclear + Grid equipment + Energy storage

The investment thesis becomes much larger than Silicon Valley. An AI data center may require billions of dollars of computing equipment, but without reliable electricity, that equipment cannot generate revenue.

7. Cooling Is Becoming Critical Infrastructure

Almost all that electricity eventually creates heat. That makes cooling another AI hard-asset theme. High-density AI servers require sophisticated thermal management.

The infrastructure can include:

Liquid cooling + Pumps + Chillers + Heat exchangers + HVAC + Water infrastructure

As server racks become more powerful, traditional air cooling can become increasingly challenging. This illustrates the second-order nature of AI investing.

The obvious trade is:

AI → Semiconductor

The deeper chain becomes:

Semiconductor → Server → Data center → Electricity → Cooling → Grid → Construction

Wall Street is increasingly examining the entire chain.

8. AI Infrastructure Can Have Higher Barriers to Entry

Launching an AI application can sometimes require relatively little upfront capital. Building a hyperscale data center is different. You need:

Land + Permits + Power agreements + Construction + Chips + Cooling + Networking + Billions in capital

These create significant barriers. Physical scarcity can therefore become economically valuable. There might be thousands of developers capable of launching another AI application.

There are far fewer locations where a massive data center can secure enough land, electricity, connectivity and regulatory approval.

Scarcity changes investment economics.

9. Infrastructure Demand Doesn’t Guarantee Investment Returns

This is where investors need caution. A booming industry doesn’t automatically make every company associated with it a good investment.

Potential risks include:

1. AI infrastructure overbuilding
2. Expensive valuations
3. Rapid semiconductor depreciation
4. Higher electricity prices
5. Grid delays
6. Interest-rate and financing risk
7. Technology becoming more efficient
8. Changes in AI demand
9. Stranded data-center capacity
10. Regulatory/environmental constraints

The IEA itself emphasizes uncertainty. Its updated analysis says bottlenecks across the AI infrastructure value chain are limiting more aggressive near-term data-center scenarios despite booming investment and enormous project pipelines.

In other words:

Demand can be real while valuations are still wrong.

AI Hard Assets Investors Are Watching

This is why the phrase “AI infrastructure trade” encompasses much more than chipmakers.

The Economics Behind the Pivot

Consider two hypothetical investments.

AI Application

An AI app needs:

$1 million startup capital

But 500 competitors can potentially build similar products.

AI Data Center

A hyperscale data center might require enormous capital, years of planning, specialized chips, grid connections and construction. Competition is constrained by physical resources.

That doesn’t automatically make the data center a better investment. It simply creates different economics:

Software advantage = innovation and distribution

Infrastructure advantage = capital + scarcity + capacity + location

Wall Street increasingly wants exposure to both.

Final Thoughts

The biggest change in AI investing is that artificial intelligence is no longer merely a software story. It is becoming an industrial infrastructure story.

The statistics are difficult to ignore.

Data-center electricity consumption increased about 17% in 2025, while AI-focused data-center consumption grew around 50%. The IEA expects global data-center electricity consumption to reach approximately 950 TWh by 2030—roughly double 2025 levels.

Meanwhile, five major technology companies spent more than $400 billion in capital expenditure during 2025, with another dramatic increase expected in 2026. That creates an investment chain extending far beyond AI apps:

AI → Chips → Servers → Data centers → Electricity → Grid infrastructure → Cooling → Fiber → Land

The crucial insight isn’t that AI applications no longer matter. It’s that every successful AI application eventually needs something physical to run on.

Software may determine who wins the AI race. But GPUs, electricity, data centers and networks determine whether that race can happen at all.

That is why the next phase of the AI investment boom is increasingly about hard assets behind artificial intelligence—not merely the apps appearing on top of them.

FAQ

Is Wall Street abandoning AI software?

No. The better description is that the AI investment theme is broadening. Software remains important, but massive infrastructure spending is drawing attention toward semiconductors, data centers, electricity and associated physical assets.

Why are data centers so important for AI?

AI models need enormous computing capacity for training and inference. The IEA says AI-focused data-center electricity consumption surged approximately 50% during 2025.

How much electricity will AI data centers need?

The IEA expects overall global data-center electricity consumption to roughly double from 485 TWh in 2025 to around 950 TWh in 2030, with AI-focused facilities growing considerably faster.

How much are technology companies investing?

The IEA says capital expenditure among five major technology companies exceeded $400 billion in 2025 and is expected to rise another 75% in 2026, primarily driven by data-center investment.

Microsoft alone now expects approximately $175 billion in calendar-year 2026 capex after adjusting for its updated treatment of certain data-center leases.

Source: finance.yahoo.com