When a company wants to build a conventional factory, it looks for land, labor, transportation, and raw materials.

But when building an “AI factory,” another factor moves to the top of the list:

Where will the electricity come from?

This problem is already changing where data centers are built.

In the UK, Drax announced a partnership to develop a large AI data center at the Drax Power Station site in northern England. The plan targets up to 100 megawatts in the first phase, starting in 2028, followed by an ambition to add 500 megawatts between 2028 and 2031, with the project potentially exceeding one gigawatt in the future. Brookfield estimated that the total potential investment, including chips, could exceed £20 billion.

The question here is not just: why this scale?

It is also:

Why build it at a power station in the first place?

The problem isn’t just generating electricity

A data center needs enormous amounts of electricity, but having electricity in a country doesn’t automatically mean it can be used anywhere.

There must be enough capacity in the transmission grid, along with substations, transformers, and connections capable of delivering that power to the site.

This is where a problem known as time to power comes into play.

A company may have the land, financing, and chips, yet wait years for an electrical connection with enough capacity.

That is why sites with existing power infrastructure are so valuable.

The Drax site, for example, has access to 4 gigawatts of grid capacity, as well as 2.6 gigawatts of operational generation capacity, in addition to land, cooling systems, and proximity to fiber-optic networks.

In other words, the project isn’t starting from scratch.

Much of the “highway” the electricity needs is already in place.

From industrial site to digital infrastructure

This reveals a fascinating economic shift.

Old industrial sites may gain new value not because of the factory that once stood there, but because of the infrastructure that grew around it.

A large power station has historically required a robust grid, extensive land, and cooling, communications, and transportation systems.

Those same features are now in demand for data centers.

So in the coming years, we may see some industrial sites shift from producing goods or electricity alone to hosting part of the digital economy’s infrastructure.

AI may seem like a digital service, but building it depends on very substantial physical assets.

Why has the grid become an economic resource?

When a data center requests 500 megawatts of electricity, it isn’t competing only with other data centers.

Factories, homes, new projects, electric vehicles, and heat pumps all need the same grid.

That is why governments are starting to consider how to prioritize strategic projects and who should bear the cost of expanding the grid.

The UK government, for example, launched the “Data Centre Covenant,” which links faster planning and access to electricity to conditions that include data centers paying their share of energy costs, using water and electricity efficiently, and providing jobs and benefits to local communities. The UK also aims to reach around 6 gigawatts of AI data center capacity by 2030.

This shows that electricity is no longer just an operating expense.

It has become a factor that determines where an entire sector can grow.

But what are the trade-offs?

Being close to a power station solves part of the problem, but raises other questions.

Higher electricity demand may require new investment in generation and grids. If clean energy production does not grow at the same pace, this could increase reliance on higher-emission sources.

That is why the project is controversial: supporters see it as a way to attract investment and build domestic computing capacity, while critics warn that the rapid growth in data centers’ energy demand could put pressure on emissions-reduction targets.

AI is redrawing the industrial map

In the old digital economy, location seemed less important: software could be developed almost anywhere.

But the AI economy is putting location back in the spotlight.

A processor needs electricity, electricity needs a grid, and the grid depends on infrastructure that took decades to build.

That is why the best plot of land for an AI center may not be the cheapest one.

It may be the plot with megawatts ready to go right next door.

The race for AI isn’t just about who has the best model or the most chips. It also comes down to a simpler question:

Who can get the electricity there first?