23rd June 2026
This is becoming one of the most important economic questions of the next decade.
A few years ago, people worried that Bitcoin would consume huge amounts of electricity. Today, the concern is shifting towards AI because AI's power requirements are growing much faster than cryptocurrency mining.
How much power does Bitcoin use?
Estimates vary, but Bitcoin mining is generally thought to consume around 100–200 TWh (terawatt-hours) of electricity per year globally.
To put that into perspective:
The UK uses roughly 250–300 TWh of electricity annually.
Bitcoin alone consumes electricity comparable to a medium-sized industrialised country.
Bitcoin's power use has been broadly stable because it depends on the economics of mining and the Bitcoin price.
How much power does AI use?
This is where things get interesting.
The International Energy Agency estimates that all data centres currently consume around 415 TWh annually, rising to about 945 TWh by 2030. AI is expected to be the biggest driver of that growth.
In other words:
Activity Electricity Use
Bitcoin mining 100–200 TWh
Global data centres today 415 TWh
Global data centres by 2030 945 TWh
AI is not just overtaking Bitcoin it is moving into an entirely different league.
Why is AI so power hungry?
Every AI query requires:
Thousands of processors working simultaneously.
Massive memory systems.
Cooling systems.
Continuous operation 24 hours a day.
Training a large AI model can consume enormous amounts of electricity, but the bigger issue is now the millions of daily users generating constant demand.
Who will pay for all this infrastructure?
Ultimately, everyone.
1. Big technology firms
Companies such as Microsoft, Amazon, Google and Meta are currently spending hundreds of billions of pounds on data centres, chips and power agreements.
Initially, shareholders bear the cost.
2. Customers
Over time those costs are passed on through:
Cloud computing fees
AI subscriptions
Software licences
Business services
Just as consumers eventually paid for broadband and mobile networks, they will ultimately pay for AI infrastructure.
3. Electricity users
This is where it becomes politically sensitive.
Grid operators will need:
New power stations
More transmission lines
More substations
More battery storage
Some of these costs will appear in electricity bills and network charges.
What happens in the UK?
This could create opportunities as well as challenges.
For example, northern Scotland has:
Abundant wind resources.
Spare land.
Cooler temperatures that reduce cooling costs.
Potential for new transmission links.
Places such as Caithness, the Highlands and parts of Aberdeenshire could become attractive locations for energy-intensive data centres if grid capacity is expanded.
That is one reason there is growing discussion about whether Scotland should use more of its renewable power locally rather than exporting it south.
The bigger question
The real issue may not be whether AI can be powered.
The IEA believes it can be, through a mixture of renewables, gas generation, nuclear power and grid expansion.
The bigger question is whether society gets enough economic benefit in return.
If AI boosts productivity significantly, the investment may prove worthwhile. If it mainly replaces existing activities while consuming vast amounts of electricity and capital, there could be growing public resistance to subsidising the infrastructure.