Finite Resources Could Cause a Data Centre Crunch in 2026

The UK needs to start with some simple steps.  Most training starts at college but a college tutor for electrical work gets paid far less than if they are working as an Electrician.  Granted the tutor get sick pay and substatial holiday and pension which a self employed house basher does not but the fact still remains.  To train the next generation of Electrician you need Electrical tutors and to get those the industry needs to pay a better salary.  

Data centres need to find better ways to convert the heat produced back into usable energy while remaining 99.999% uptime.  99.999% uptime (known as "five nines") means a system is perfectly operational 99.999% of the time. Over a full 365-day year, this allows for a total of only 5 minutes and 15 seconds of downtime

Data centres need to find better ways to convert the heat produced back into usable energy

Well there is at least one instance of a new data centre which intends feeding its excess heat into a district heating scheme - https://www.bbc.co.uk/news/articles/cd7p21599zgo

The heat "energy centre" is already under construction - https://1energy.uk/networks/bradford-energy-network/  

- Andy.

That’s a great example. Reusing waste heat from data centres through district heating schemes could help address one of the sector’s biggest challenge energy efficiency.

Projects like the Bradford Energy Network show how data centres can contribute to local communities by supplying excess heat to homes and businesses, rather than letting it go to waste. As demand for data centres grows, I think more developments should incorporate this type of circular energy approach alongside renewable power generation and grid upgrades.

Innovations like these will be important if the UK is to expand its digital infrastructure in a sustainable way.

They could also wrap the data centre in solar panels to generate energy that could maybe be used to power some of the electric vehicle or connect that with the heat generated and then employ some kind of steam generator to make energy.  

The possibilities are out there, we just need the UK industry and government to look towards engineers for solutions.  These solutions should then be given some kind of tax incentive as data centres become more common place on the UK landscape

The proposed power densities are pretty extreme https://www.datacenters.com/news/ai-cooling-systems-must-support-100-kw-racks   I'm not sure that a coating of solar panels generating a few hundred watts per panel on a good day will be enough to make a dent in it.  
Lets take that 100kW per rack figure for a moment - that's the same sort of power as the transformer that supplies 25 house on each phase in a small village. For just one rack.

So, if we need to unhook a small village from the grid to supply each rack, how many fully populated towns have to go dark to power just one bit-barn full of them ? !  Is that a price we are prepared to pay for dancing cat videos and anodyne summaries of documents we dont want to read? 

Something is simply not going to happen at the scale, speed or power density people imagine, and I don't think the fraction of a giga-watt data centres will ever materialize.
What you may get is a new power station, for which the base-load is assumed to be the data-centre.

 That's not to say we should not be looking at energy recovery, district heating and so on, we absolutely should, for all new power stations.
I just  caution that there is a lot of hype and really it is quite nuanced, and we don't want a town to lose heat due to a change of profitability of some business model that may not be right.
Mike.

Data centres are going to become more common in the UK and currently they have several issues

Planning permission

Site cooling


Making the site cooling feed into district heating could turn an issue into a revenue stream for the data center.  District heating systems would also help with planning permission


As an example this morning on Sky newse there was a piece about how a data centre was using water from the Thames to cool itself so they are currently through away the heat energy.  

www.youtube.com/watch

It also raised the question of how mush water was being used and if this could affect water levels during the summer months when excess heat build up in the data center was at peak and during the same time that people in the local area may be facing water shortages.


Looking to the future are we going to be discussing climate change/global warming (Personally I prefer the term climate change) being caused by data centres rather than from fossil fuels

Certainly running hot water to waste into a river or similar to the extent that it affects the wildlife is not the sort of thing that should really be contemplated without full consideration of better alternatives.

At the moment of course if the data centres are anything like as thirsty or as numerous as projected, we will double our rate of fossil fuel use, or if you prefer to thing about it the other way, halve the time until it runs out, or more accurately becomes uneconomically rare to waste just for energy.

Its not an appealing or realistic prospect, and in reality the change in national power consumption will have to be less. so there will have to be fewer bit barns built, or there will be less thirsty equipment within. What I think is  more likely is that as chip designs improve the same computing power will consume less, and the technology will only really take off in a big way, once the energy requirement per process, has fallen to a more manageable level. Until that point only the big money projects will go ahead.

Mike 

You raise some valid concerns, particularly around energy consumption, water use and the environmental impact of future data centres.

I agree that cooling and power demand will be major challenges. That’s why I think future data centre development needs to be linked to investment in renewable energy, grid upgrades and more efficient technologies. The example of using river water for cooling also highlights the need for careful management of local resources and consideration of alternative approaches such as heat recovery and district heating.

On the point about energy consumption, I suspect technology will evolve in both directions. Demand for computing power will continue to increase, but advances in chip design, cooling systems and energy efficiency should help reduce the power required per unit of computing. Historically, computing has become far more efficient over time, even as overall usage has grown.

Rather than limiting growth, I think the challenge is ensuring that data centre expansion is planned sustainably so that economic benefits, technological innovation and environmental responsibilities are balanced effectively.

It will certainly be an interesting debate over the next decade as AI and digital infrastructure continue to expand.

You raise some valid concerns, particularly around energy consumption, water use and the environmental impact of future data centres.

I agree that cooling and power demand will be major challenges. That’s why I think future data centre development needs to be linked to investment in renewable energy, grid upgrades and more efficient technologies. The example of using river water for cooling also highlights the need for careful management of local resources and consideration of alternative approaches such as heat recovery and district heating.

On the point about energy consumption, I suspect technology will evolve in both directions. Demand for computing power will continue to increase, but advances in chip design, cooling systems and energy efficiency should help reduce the power required per unit of computing. Historically, computing has become far more efficient over time, even as overall usage has grown.

Rather than limiting growth, I think the challenge is ensuring that data centre expansion is planned sustainably so that economic benefits, technological innovation and environmental responsibilities are balanced effectively.

It will certainly be an interesting debate over the next decade as AI and digital infrastructure continue to expand.

Terry osifo said:

but advances in chip design, cooling systems and energy efficiency should help reduce the power required per unit of computing.

Have you seen this: https://www.bcs.org/articles-opinion-and-research/brain-inspired-chip-material-promises-low-energy-use-ai/ ?

   - Andy.

Yes, that’s a very interesting example. The Cambridge work on neuromorphic (brain-inspired) computing highlights exactly the sort of developments I had in mind. By combining memory and processing in the same device, these memristor-based architectures can significantly reduce the energy wasted moving data between separate components, which is one of the major contributors to AI power consumption. The researchers suggest such approaches could reduce energy use substantially compared with conventional architectures. (BCS)

While these technologies are still emerging and not yet deployed at scale, they illustrate how advances in chip design, materials science, and computing architectures could improve AI’s energy efficiency over time. Alongside improvements in cooling systems and renewable energy integration, they may help offset some of the increasing power demands of large-scale AI workloads.

Thanks for sharing the article.

Best regards,
Terry

One of the aspects I find particularly interesting is the potential impact on the UK. If technologies such as neuromorphic chips can be scaled successfully, they could help address concerns about the growing energy demands of AI and data centres.

For the UK, this could support both economic growth and sustainability goals by enabling more efficient computing infrastructure, reducing pressure on the power grid, and strengthening the country’s position in AI and advanced semiconductor research. Given the UK’s strong academic research base, there may also be opportunities to attract investment and create high-skilled jobs in emerging computing technologies.

It will be interesting to see how quickly these concepts move from the laboratory into commercial applications.

Indeed Terry,

It occurred to planners in China that by building data centres under-sea - on the sea bed - 90% of the electricity could go into processing.

And storing so much data, such as images of my electricity meter reading - worthless data. 

However, whilst this might be heresy in the IET, I think a combination of:

• Unfulfilled RoI an AI/AGI/AMI;
• Unfulfilled data centre RoI;
• Risks of circular investment, the debt burden and securing funding;
• Resource crisis (especially copper, the key to everything else, even thorium reactors)
• Oil, as I've said, or more specifically and immediately, CONTROL of the oil, East-West;
• Other critical metals, minerals and more geopolitics. 

I think these will determine much more selective use of AI and data centres and that could be a very good thing. 

The Technosphere is very good at sub-specialisation driving the concept to the point of unviability, or unproductive investment; in simple terms, missing the point.

I'd point to HS2. If only the technical, civil engineering, land purchase, spiralling cost (and even testing) impacts of the quest for the fastest line speed had been admitted earlier, we might have a longer railway that answered the original brief of adding capacity.

Best

Chris