DEVELOPERS BEHIND Scotland’s proposed new generation of huge AI data centres say the developments could create thousands of jobs – but what would those jobs actually be?
Billions of pounds of investment is being proposed in hyperscale data centres across Scotland, with employment among the major economic benefits put forward by developers.
And there is little doubt that constructing facilities of that scale creates significant amounts of work.
Data centres also need engineers, IT technicians, electricians, security teams and maintenance staff once they are operating.
But the headline number of jobs “supported” by a development can be very different from the number of people ultimately turning up to work at the data centre itself.
That is because economic assessments can count several layers of employment: temporary construction work, permanent on-site staff, jobs in companies supplying the facility, and jobs elsewhere in the economy supported when those businesses and workers spend their money.
The proposed Larbert Data Centre Campus provides a useful example of how those figures are calculated.
Developer Apatura is seeking permission for a 300MW data centre at Glenbervie Business Park near Larbert, with more than £2billion of investment expected if it goes ahead.
Its economic assessment, carried out by consultants Glic, estimates the development could generate 10,340 job-years across Scotland during construction and support 1,382 jobs annually once operational.
Apatura chief executive Giles Hanglin said the figures were calculated using Scottish Government economic tables, national accounts and official employment data.
He said the Larbert campus would bring “more than 10,000 net job years during construction”, of which nearly 2,000 could be fulfilled locally.
But neither figure means 10,000 people would be working simultaneously on the development.
The biggest jobs boost comes while they are being built
Constructing a hyperscale data centre is an enormous infrastructure project.
The buildings themselves have to be erected, but developers also require electrical infrastructure, substations, cooling equipment, fibre connections and specialist engineering.
That creates work for builders, electricians, mechanical and electrical engineers, equipment suppliers, architects and other professional services.
In Larbert, Glic estimates construction would produce 6,034 direct job-years in Scotland, with another 4,306 indirect and induced job-years, giving the 10,340 total.
In Falkirk itself, the equivalent forecast is much smaller: 1,191 direct and 638 indirect and induced job-years, or 1,829 in total.
A job-year is an economic measure equivalent to one full-time job lasting for one year.
So 10,340 job-years does not mean 10,340 individual vacancies.
Glic assumes construction would be spread over four years and says, on that basis, the project would sustain the equivalent of around 2,585 full-time jobs a year across Scotland during the build.
The distinction between construction and operational employment is not unique to Larbert.
A major investigation by Virginia’s Joint Legislative Audit and Review Commission found the principal economic benefit from data centres there came during construction.
Representatives told the commission that a typical 250,000 sq ft data centre could have around 1,500 workers on site at the height of construction, but around 50 full-time workers once operating, approximately half of whom could be contractors. The American facilities are not directly comparable with the much larger Larbert proposal, but the findings illustrate the labour-intensive nature of building data centres compared with running them.
What happens when the servers are switched on?
Data centres are highly automated, but they are not unmanned.
Servers and networks need IT staff, while the enormous electrical and cooling systems required to keep them operating around the clock need engineers and technicians.
There are also jobs in security, facilities management, cleaning, repairs and maintenance.
For Larbert, the difference between those jobs and the overall headline figure is significant.
Glic estimates 1,382 long-term jobs would be supported across Scotland, including 515 within the Falkirk area.
But 297 are forecast to actually be based at the Larbert site.
The other 218 Falkirk jobs would be supported through local suppliers and spending, while the Scotland-wide assessment includes a further 1,085 indirect and induced jobs.
Apatura says this broader employment is an important part of the development’s economic impact.
Hanglin said the local workforce would extend beyond computer-based roles to engineers, electricians, maintenance workers, security staff and building services, with an estimated average salary of almost £60,000 for the 297 on-site positions.
There is another distinction within those 297 jobs.
Glic estimates 101 would be IT technicians and other staff employed directly by the eventual data centre user or users.
The remaining on-site employment includes areas such as facilities management, maintenance and security.
Its methodology assumes Apatura would subcontract those services rather than directly employ the workers itself, but counts the employment as “direct” because the work would take place at the site.
That means “direct jobs” in an economic impact assessment does not necessarily mean people directly employed by the data centre operator.
So where do the other jobs come from?
The next layer is the supply chain.
A working data centre needs electricity, connectivity, maintenance, repairs, security and other goods and services.
Businesses providing those services employ people because of the money being spent by the data centre.
Economists describe these as indirect jobs.
Then there are induced jobs.
An engineer, security worker or supplier earning money from the development will spend some of their wages elsewhere – in shops, restaurants, leisure businesses and other parts of the economy.
Economic modelling estimates how much additional employment that spending supports.
So a person counted within the economic impact of a data centre does not necessarily work for a data centre company – or even within the technology industry.
This type of modelling is widely used to assess major investments.
Glic used Scottish Government Input-Output tables, national accounts and employment statistics, and also made adjustments for “deadweight”, economic activity that could have happened anyway; “displacement”, where new activity replaces activity elsewhere; and “leakage”, where spending leaves the area being studied.
In fact, its assumptions recognise that a considerable amount of spending associated with a global data centre operator may never reach Scottish businesses.
For some indirect spending by the eventual data centre user, it applies leakage assumptions of 90 per cent at Falkirk level and 75 per cent at Scotland level, partly because the future users are likely to be headquartered elsewhere and could procure services such as software, insurance and financial services outside Scotland or internally.
How many people does an AI data centre actually need?
There is no single answer.
Different facilities operate in different ways, and employment can vary considerably depending on whether they are hyperscale centres operated by major technology companies, colocation facilities hosting equipment for multiple customers, or more lightly staffed “dark” sites.
Glic itself acknowledges that uncertainty.
Its estimate of 101 data-centre-user jobs at Larbert is not based on a final staffing plan from a named future operator.
Instead, it uses an employment-density assumption of 440 square metres per worker.
The evidence cited in the report ranges from around 180 sq m per worker in a colocation data centre to 1,400 sq m in a wholesale dark site, demonstrating just how widely employment intensity can differ.
There is evidence, however, that large data centres can produce employment outside the building itself.
Researchers at Brookings examined around 1,500 US facilities and compared areas receiving data centres with places where proposed facilities had been cancelled.
They found the arrival of a first large data centre increased employment in data processing and, in the case of hyperscale facilities, telecommunications.
For a typical county, the researchers estimated an overall effect of around 100 to 200 jobs, depending on the type of facility. They also stressed that the US findings should not be treated as a simple headcount of staff inside a data centre.
The UK industry argues the potential employment effect could become substantial as the sector grows.
Trade body techUK estimates operational UK data centres currently support around 43,500 jobs across the economy, and has projected that faster expansion of capacity could create 40,200 additional direct operational roles by 2035.
Those projections are disputed. Environmental economics think tank Verdant argued this month that actual direct employment at existing and proposed UK facilities suggests significantly fewer jobs than the industry estimates.
The difference demonstrates why definitions matter when data centre employment figures are discussed.
Are the headline job figures wrong?
Not necessarily.
A job supported in a supply business is still a job. A security worker employed by a contractor still has a job at the data centre. And large construction projects can sustain significant employment even though those roles disappear when construction finishes.
But “jobs supported”, “construction job-years”, “local jobs” and “people permanently working inside a data centre” are not interchangeable figures.
Larbert demonstrates that particularly clearly.
Its economic assessment forecasts 10,340 construction job-years across Scotland, of which 1,829 are expected in Falkirk.
Once operating, it forecasts 1,382 jobs supported across Scotland, 515 in Falkirk, 297 physically based at the data centre and around 101 employed directly by its eventual user or users.
The jobs case for Scotland’s new data centres therefore goes considerably wider than the people working among the servers.
And understanding that distinction is key to working out what the thousands of jobs attached to Scotland’s proposed AI data centre boom would actually look like.
Iona Craig has an NCTJ from Glasgow Clyde, PhD in English Research from University of Strathclyde and a Masters Distinction in History from the University of Stirling. Currently working as a Journalist intern at Deadline News.