Finite Resources Could Cause a Data Centre Crunch in 2026

I thought the purpose of these data centres was to reduce jobs, not create them.

I think it’s less about reducing jobs and more about changing them. Data centres still require skilled engineers, technicians, cybersecurity professionals and infrastructure specialists. The real challenge for the UK is making sure people have access to the training needed for these evolving technical roles.

Interesting thoughts, but a couple of observations. 

There are some very unrealistic estimates of future energy demand or perhaps supply. 
This article for example

https://eandt.theiet.org/2026/05/01/finite-resources-could-cause-data-centre-crunch-2026 

 Cites  "UK energy regulator Ofgem has warned that 140 proposed UK data centres could add 50GW (5GW more than current peak UK demand) and that satisfying such demand could have an impact on domestic costs."
Well yes.  

A few seconds of critical thought tells us that won't happen, well not on any sensible timescale "could have an impact" is just so much bubbles, - to do that for real, we'd only need a second British Isles to decant all the people to as the lights will have gone off in this one !

So perhaps the more immediate question should be framed another way - perhaps "how much AI centric data centre stuff can we actually sensibly build and run in the UK in the short term without putting blowing the main fuse?" as it were,
 and then design our software footprint to fit within that capability limit, and that probably includes imposing hard geographical constraints as to where the available power is.

I don't disagree with uprating the grid (as in some cases just replacing the rustier bits of what we already have is already a pressing urgency) and designing in more renewable capacity are desired, but that’s needed with or without AI, and is required just  to maintain us in the lifestyle to which we are accustomed in the coming  post gas and oil future.

Much of the big grid stuff is post war but pre 1970, and was really built to answer a different need, but if we assume we can replace it all in a similar timescale, then we should assume what we design now will be in service well past 2060, maybe wuite a bit beyond 2100, and then that really needs us to think through and out the other side of  the current obsessions with both fossil fuels and bits of silicon.

I'm also not too sure what sort of 'engineering apprentices' you would want or need for this future - current schemes already seem to struggle to find the right sort of applicants. Also just creating apprentices is a good start, but not really how you develop a workforce, more how you produce the seeds for possibly having one. If there is not then the secure career and knowledge development path, all the way up to the decision making layers, folk will then leave and go abroad or be lost to management or marketing or something, and you lose the design authority skills for the next generation of projects. So before you make too many starters, make sure there is a technical roadmap all the way to retirement.

 Mike

The main thing we need in the UK is sovereign cloud infrastructure that's not owned or run by American companies. I think European owners could be trusted, but personal details of UK citizens should not be controlled by American companies, as they can't be trusted because of US laws.

I'm inclined to agree, far too few people, even those advising the government, seem to be aware of the implications of the US cloud act - which as this article  summarises,  among a lot of far more innocuous stuff, also means all US based firms like Microsoft, Google and most of the rest of the 'big' office suites, can be obliged to hand over any data they have to US authorities, and not just that held on servers in the US and to do so without telling the organisations whose data is involved, so it may already be happening. At least as Microsoft lawyers are keen to point out,  such data extraction does need a suitably framed US court order first, which is something, but I can well imagine a situation, where it is not as  reassuring as it is intended to be.

More generally I don't really understand the rush to store sensitive company and personal data on any hardware not under some local control, and certainly not hardware that is one US court order away from being totally out of the country.  
If we must have off-site storage and make ourselves vulnerable to attacks on the network, at least keep it on a landmass and in the hands of organisations governed by the same jurisdiction.
The AI thing just adds another layer of complexity.
Mike.

I suspect the ever increasing demand for AI processing will be contained once the big firms stop giving the service away for free and start charging a realistic amounts. At the moment they seem to be chucking AI at everything (including every Google search, whether you wanted it or not) - presumably trying to create a market that they can capitalise on in the future. The venture capitalists won't keep that up for ever though. Once end users start having to pay the real costs, I'm sure many will find that the simpler and cheaper (if "old fashioned") solutions are still perfectly usable or that they didn't really need to fabricate another picture with an extra silly hat on it.

   - Andy.

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