Texas Just Pressed Pause on AI. Here’s Why That Matters More Than the Next ChatGPT.

For the past two years we’ve been told that the race for AI supremacy would be decided by better models, faster chips and bigger investments.

Texas has just reminded us that’s only half the story.

The State of Texas has paused approvals for new AI data centres requiring connections to its electricity grid while regulators audit hundreds of proposed projects. At first glance it sounds like a local planning issue.

It isn’t.

It’s one of the clearest signals yet that the AI revolution is running into the physical limits of the real world.

The AI boom has a power problem

Much of the conversation around AI has focused on software.

Large language models.

Agentic AI.

Autonomous systems.

But none of those exist without enormous amounts of electricity.

Texas estimates that around 90% of the 474 gigawatts of new electricity demand currently under review comes from proposed data centres. That’s more than five times the state’s current peak electricity demand. State regulators now want developers to justify their energy requirements, water consumption, ownership structures, tax incentives and local community impact before additional approvals are granted. (Reuters)

Suddenly, the limiting factor isn’t intelligence.

It’s infrastructure.

The next AI race isn’t about algorithms

For years the technology industry has competed on computing power.

The next competition will be for:

• Electricity
• Grid connections
• Land
• Water
• Cooling
• Skilled construction workers
• Community support
• Project finance

In other words, AI is becoming an infrastructure industry.

That changes everything.

Wall Street has noticed

Reuters reported this week that banks are becoming much more selective when financing AI data centres.

Instead of simply asking whether a hyperscaler has signed a lease, lenders are now examining:

• Can the local grid actually supply the power?
• Is there enough water?
• Will planning permission survive local opposition?
• Could political sentiment change?
• Will the project still make sense if AI demand slows?

These are questions normally associated with airports, power stations and transport infrastructure.

Not software companies.

Local communities are becoming a strategic risk

Perhaps the biggest surprise has been where the resistance is coming from.

Many rural communities initially welcomed data centres as symbols of investment and technological progress.

Now they’re asking much harder questions.

How much water will they consume?

Will electricity prices increase?

Where are the permanent jobs?

Who benefits?

What happens to farmland?

Across the United States, opposition is growing as communities weigh promised economic benefits against energy consumption, noise, land use and environmental impact. Texas is no longer an isolated case.

Technology companies are discovering something property developers have known for decades:

Planning permission is never automatic.

The real winners may not be AI companies

Ironically, the biggest beneficiaries may not be the companies building AI models.

Instead, value is increasingly shifting towards businesses that can solve the infrastructure challenge.

Grid operators.

Power-generation companies.

Battery storage.

Cooling technologies.

Electrical engineering.

Construction.

Project finance.

Energy management software.

The AI economy is quietly creating an entirely new industrial ecosystem.

What this means for banks

Having spent years leading strategic transformation in banking, I find this part particularly interesting.

Banks are no longer simply financing technology companies.

They’re financing digital infrastructure.

Every proposed AI campus now resembles a major infrastructure project requiring specialists in project finance, energy, environmental risk, structured lending and long-term operational analysis.

That creates opportunities for institutions with expertise in infrastructure finance and cross-border capital markets.

It also introduces new risks.

If AI demand fails to justify today’s extraordinary investment assumptions, lenders could find themselves financing some very expensive buildings filled with rapidly depreciating technology.

The Lomond Logic view

The Texas decision is not an anti-AI story.

It’s the first serious reminder that digital revolutions still depend on physical reality.

For years we’ve talked about artificial intelligence as though it existed in the cloud.

It doesn’t.

It lives in warehouses full of servers, powered by power stations, connected by transmission lines, cooled by millions of litres of water and financed by billions of dollars of capital.

The next chapter of AI won’t be written solely by software engineers.

It will be written by electrical engineers, energy companies, planners, financiers and communities deciding whether these facilities should be built at all.

That’s a much bigger story than the next chatbot.