Lanarkshire AI Growth Zone will create a tiny fraction of the jobs promised

We have had a response back from a Freedom of Information Request, published in the Guardian today, that we sent to the UK Government regarding the Jobs estimates for the AI Growth Zone in Lanarkshire. It shows that the UK Government expect less than 10% of their headline figure to be jobs delivered by the AI Growth zone.

It turns out that only 10% of the figure they gave are for jobs at the data centre – and that, itself, is an estimate taken from a  developer in Northumberland and not from real figures. When we look at real figures from data centres, as we did in our recent report, Employment and Data Centres, actual figures of employment in data centres are much, much lower, between 20 and 50 jobs.

Incredibly, the UK government estimated the number of jobs that the AI growth zone in Lanarkshire would create by taking an industry estimate from the Cambois proposal in Northumberland and multiplying it down for the size of the Chapelhall site. This lead to a figure of around 300 jobs – less than 10% of their headline figure. It seems strange to take estimates from the developer of another site (the figures, incidentally, had already been debunked in the Financial Times).

The answer to the FoI

The answer that the UK Government gave when asked how they had calculated the 3400 figure for jobs created by the AI Growth zone was “The “more than 3,400 jobs” estimate is based on conservatively scaling projected jobs using evidence from the Cambois data centre development. This is expected to directly create 1,200 construction jobs and 400 operational roles, and adjusting these figures for the projected capacity of the Lanarkshire AI Growth Zone. Conservative employment multipliers are applied: approximately 1.5-1.6 across construction and operations roles.“

The calculation

When you scale the figures for the Cambios data centre using the scaling factor the Government used (5/6) down to the 500MW Lanarkshire AI growth zone and use a figure of 1.55 (the average of the figures given), you come up with 333 jobs in the data centre, 1000 construction jobs and the rest (2066) being induced jobs

So way more than half the figures (61%) in the estimate come from induced jobs (multiplying up for how many roles are created in the wider economy.)

In our report on Employment and Data Centres we showed that developer estimates on data centre jobs tend to be inflated in relation to actual figures from operating data centres.

Our conclusion is that the UK Government figures given in their press release could be a hundred times bigger than the probable jobs that the data centre will create.

Not only this, but our report explains how data centres crowd-out development which is far more beneficial to communities, this is because they use so much land, and that land that then can’t be used for other, more productive, purposes. Their vast energy use is also problematic because this can prevent other business, that would provide more jobs, or housing developers building homes, cannot get connections to the grid. 

Temporary construction jobs

It’s also unclear how the figure of more than 1000 construction jobs is estimated. These are obviously temporary, but there are also unlikely to be 1000 people on a building site for the entirety of the 12-18 months of the construction. Different trades are there at different times. How is the very temporary nature of these jobs accounted for in figures – this is not at all clear and the UK Government needs to put flesh on the bones of these figures. In addition these are evidently not new jobs, they are roles that will be building other structures at other times, locally in many cases but, for specialist roles, they will be brought in from outside. 

The UK Government has also used a multiplier on construction jobs, even though the construction jobs aren’t new jobs created just for the year that the data centre takes to build. They are coming from other areas of the construction economy and, if they weren’t working on the data centre, they would be working on another construction project.

These indirect jobs are also not guaranteed to be in Lanarkshire, or even in Scotland. The UK Gov said, in answer to a Parliamentary Question in the Commons, “We do not make specific assumptions about the nature or geographical nature of jobs indirectly related to AI Growth Zones”.

The UK Government’s Response to our Jobs and Data Centres Report

In the UK Government’s response to our report on Jobs and Data Centres they admitted that data centres do not create many jobs. The spokesperson said: “We have never claimed that thousands of people will work inside a data centre day to day. Our figures include construction, supply chain and wider jobs across the economy.”

However the UK Government press release announcing the AI Growth Zone in Lanarkshire gives the headline figures and does not divide out figures for construction, indirect and induced jobs. They know that, when they give figures for jobs created at the AI Growth zone with a headline ‘More than 3,400 jobs and targeted support for local communities to help tackle the cost of living as Lanarkshire named latest AI Growth Zone’, then everyone assumes that this is the number of jobs. 

It is unreasonable to expect people to realise that less than 10% of the headline figure is actually the jobs at the data centre itself, and even this figure is an industry estimate many times larger than actual employment figures from operating data centres.

The UK Government reaction to our Jobs report goes on to say

“AI data centres are enabling infrastructure, like energy networks or transport. Judging them solely on permanent on-site headcount ignores the investment, capability and job opportunities they unlock. The alternative is to send that growth and those jobs overseas, while buying back the compute from abroad.”

Local Concerns 

Local people are concerned about the extent of the development and the impacts on their green belt. Members of Woodhall Faskine and Palacecraig Conservation Group have been campaigning for a decade to protect their green belt which lies between Calderbank, Chapelhall and Airdrie.

You can read about their campaign including an interview on Out of Doors and about the recent refusal of the developer’s appeal. 

Diane Davidson of Woodhall Faskine and Palacecraig Conservation Group said:

“The 3,400 jobs promised for Lanarkshire are a fantasy. We know now that 90% of these roles are temporary construction or ghost jobs that might not even be in Scotland, or the area. To sacrifice hundreds, maybe thousands of acres of working farmland for a facility that typically employs fewer than 50 people is a betrayal.

“In the 2018 ‘Europark’ 18/000890/PPP case, we proved this greenbelt is worth protecting, yet now our farmers are being placed under immense pressure to ‘cash in’ to developers. When land value leaps from £9,000 to over £40,000 an acre with a single planning change of use. This isn’t economic growth, it’s an incentive to destroy our food security and wildlife corridors.

“This project guts North Lanarkshire’s biodiversity policies. It replaces vital carbon-sinking soil and protected habitats with a ghost grid of private wires and server warehouses. The people of Chapelhall and Airdrie bear the environmental costs, the drain on our grid and the loss of our landscape, while the data and the profits flow straight to US tech giants. We fought for our greenbelt before and won, we won’t let it be paved over for a handful of security guards and a server farm.”

Kat Jones said:

“As far as we can work out 90% of the jobs the UK Government announced for the Lanarkshire AI growth zone are either construction jobs or induced jobs, which means jobs in the wider economy (and these don’t even need to be in Lanarkshire) [1].  Even the figure of operational data centre jobs was taken from an estimate made by a data centre developer in Northumberland rather than evidence from actual data centre employment.

“It is clear that little, if any, serious economic analysis has gone into this estimate at all. The people of Chapelhall and Airdrie deserve better if they are being expected to bear the environmental, electricity grid and potential air-quality impacts of this giant data centre. 

“Our recent report on jobs and data centres looked at reported and audited figures for data centre employment globally and found that, in general, 20-50 people can be expected to work in a data centre.  We also found that data centres can crowd out development that has far more benefit to the economy, and creates far more jobs.[2] 

“Hyperscale data centres famously employ extremely few people, and have enormous impacts on the environment and communities, so we should expect much more scrutiny of job projections. The UK Government will need to find a more plausible reason to legitimise the sheer quantities of energy, land and water that will be diverted from the Scottish economy at large, to the purpose of training US big tech’s AI models.”

“If the UK Government really wanted to create tech employment in Scotland, and benefits across our whole knowledge sector, they would ditch the hyperscale AI data centres, and invest in a proper research-scale supercomputer and academic collaborations, like that at Edinburgh University, which is set to be the largest publicly-funded data centre in the UK. 

“The fact that £1.5billion investment from US AI tech giant CoreWeave into, what is now proposed as Scotland’s AI growth zone, was announced on the day of Trump’s State visit to the UK tells you all you need to know about who benefits from this deal.”

[1] the UK Gov said in answer to a westminster PQ ‘ We do not make specific assumptions about the nature or geographical nature of jobs indirectly related to AI Growth Zones’  

[2] see section in APRS Report on jobs and data centres on opportunity costs for employment. It includes impacts on house building, other industries that need the energy, components and land. 

Doubts about the investment into the AI Growth Zone

In addition, we are concerned about the collaboration between data vita and coreweave, the US AI compute company that pledged some of the money to build the Chapelhall 500MW data centre. 

This funding arrangement was announced on the day of the Trump state visit in a press release with a quote from Keir Starmer. The state dinner had guests including many from the US tech industry, and after which the US and UK announced their Tech Deal which they said would bring huge amounts of investment. Most of this seems to be investment in data centres from the list here

The deal was later paused by the US – but it seems much of the data centre spending is going ahead because, of course,  that benefits the US and they don’t have an incentive to drop out of that. 

Coreweave have pledged to put in £1.5bn into the data centre but an article from FT last week gives an estimate of the costs of a hyperscale data centre. The Financial Times estimate that a 1000MW hyperscale data centre will spend $35bn on chips and hardware, and $20bn on land, power and other inputs for a gigawatt of data. This means a 500MW data centre like that planned for Chapelhall will cost $27billion. 

With Coreweave putting in only £1.5bn there is a very significant funding gap. An article in the Guardian suggested that Coreweave is having difficulties delivering on its pledges. 

Calculation of the figures the Uk Government are using

I tried to work out what figures they were using. If they were using a 1.55 multiplier (average of 1.5 and 1.6 as they note in the FoI) then there must be 1333 jobs split between construction and operations in the data centre. And if this is on the same ratio as the jobs they estimated at Cambois then that is 1000 construction and 333 in the data centre. But they said there would be 800 high paying jobs. So I can’t imagine where they all come from – unless they are in the multiplier added. Obviously 333 is an unrealistic number from what we found in our jobs report anyway.

All in all this seems an extremely unprofessional way to estimate jobs from something that will need £10billion investment.