One Model Assumes 8.5 Jobs Per Megawatt. Twenty Planning Files Average 1.2.
AI-generated. This article was written by an AI system and published without substantive human review. Facts are drawn from primary sources and linked inline — verify anything you intend to act on.EU AI Act Art. 50(4)
The usual subject of this site is a claim whose own source document counts something narrower than the claim. This one is a variant worth more attention: two organisations, one metric, and a seven-fold gap in what they plug into it. Both published their working. Both are downloadable. And the government, defending the larger number, has ended up calling the underlying method deeply misleading.
The industry figure for the UK data centre build-out is 40,200 new direct operational jobs by 2035. The think tank figure, built from planning applications, is about 10,400.
Neither side is using a different method. They are using the same method with different multipliers, and the multiplier is where the entire disagreement lives.
What was published, and when
On 9 September 2026, Verdant, a UK economic think tank, publishedThe Jobs Mirage of Data Centres, a policy briefing by James Meadway. CNBC covered itthe same day, and carried responses from both techUK and the UK government.
The claim being challenged comes from techUK and Henham Strategy,Foundations for the Future: how data centres can supercharge UK economic growth, published November 2024. Verdant cites it at page 19.
Verdant summarises the existing-estate dispute in one sentence, and it is the cleanest statement of the gap:
"techUK claim that 43,500 jobs are already supported by the existing data centre estate. But only 24,300 of those are actually directly in data centres, the remainder being in ‘indirect’ employment"
Against that 24,300, Verdant's own bottom-up count comes to "around 4,400 direct jobs today". Applying the same re-estimated ratio forward gives "10,400 nationally, or about a quarter of the industry's projection of 40,200 for the same period."
The number that is doing the work is 8.5
Strip away the headline totals and the dispute reduces to a single multiplier. Verdant is explicit that it is not rejecting techUK's method:
"TechUK uses a jobs per megawatt (jobs/MW) formula to estimate the numbers of jobs to be created by data centres, which is used widely in planning decisions. Following techUK's method, of assuming each additional megawatt of processing power consistently creates new jobs, we re-estimate this jobs/MW multiplier."
That characterisation of techUK's method is Verdant's, and this post treats it as Verdant's claim rather than an established fact. The techUK report was not read line by line here.
Then the comparison itself:
"Across 20 proposed schemes for which we could match employment to capacity, the capacity-weighted average is 1.2 operational jobs per megawatt. However, techUK's model assumes 8.5, seven times higher, a significant over-estimate."
Verdant then does something that most advocacy research does not, and it deserves credit for it. It immediately narrows its own claim: "That is a comparison of ratios at constant site size. But if we take account of the size of the schemes actually proposed, the same evidence gives 1.6 jobs per megawatt across the pipeline."
Three ratios, three different objects
Verdant publishes three jobs-per-megawatt figures in the space of two pages, and anyone quoting this report is going to merge them. They are not the same measurement.
| Figure | What it describes |
|---|---|
| 1.2 | Capacity-weighted average across 20 proposed schemes, at constant site size |
| 1.6 | The same evidence, adjusted for the size of the schemes actually proposed |
| 8.6 | The existing UK estate on Verdant's own headcount estimate: 514 MW average load, 4,397 direct jobs |
That last row is the one that makes the argument coherent rather than merely adversarial. On Verdant's own numbers, the ratio for data centres already running is 8.6, which is essentially techUK's 8.5. The disagreement is not about what today's estate employs per megawatt. It is about whether that ratio survives contact with a pipeline dominated by hyperscale sites. Verdant's executive summary makes exactly that argument: the techUK figure "does not account for the fact that majority of new UK capacity planned is in ‘hyperscale’ data centres, which (because they operate at such a massive scale) create fewer jobs for each megawatt of computing capacity installed."
So the real claim is narrower and more interesting than "the industry is wrong". It is that a multiplier derived from yesterday's mixed estate is being applied to tomorrow's very large sites, and that staffing does not scale with capacity.
The comparison the government objected to
Verdant sets data centres against other electricity-hungry things, on the same page, using the same two columns. The table is headed "Electricity usage and jobs created for major UK installations, current estate":
| Installation | Average load (MW) | Direct employment | Jobs per MW |
|---|---|---|---|
| Data centres, on Verdant's estimate | 514 | 4,397 | 8.6 |
| Data centres, on techUK's claim | 514 | 24,300 | 47.3 |
| Steelworks (iron and steel) | 220 | 64,000 | 291 |
| Car plants (motor vehicles) | 350 | 140,000 | 400 |
| Hospitals and health facilities | 755 | 2,737,000 | 3,626 |
Read the second row before drawing conclusions from the rest. Even on techUK's own headcount claim, at 47.3 jobs per megawatt, data centres sit an order of magnitude below steelworks and two below hospitals. The industry number and the think tank number land on the same side of that comparison. They differ by a factor of five and the next category up is six times higher again.
Both rebuttals, in full
techUK told CNBC:
"Verdant offers its own employment estimate using an unreliably small sample size. Furthermore, the report reduces the economic value of the sector to only the number of people physically working inside data centres, not the countless jobs that are supported by data centres and the services they provide."
A UK government spokesperson told the same outlet:
"This report relies on a deeply misleading jobs-per-megawatt metric, makes false comparisons with hospitals and schools, and appears to confuse electricity capacity with actual consumption." And: "it discounts the wider economic value of compute and the national security and sovereignty case for building critical digital infrastructure in Britain."
Take those seriously, because two of the objections are substantive. Twenty schemes is a small sample for a national projection, and Verdant does not claim otherwise. Capacity and consumption genuinely are different quantities, and a booked megawatt is not a consumed megawatt.
But notice what neither statement contains. techUK's response disputes Verdant's sample size and Verdant's scope. It does not say that the jobs-per-megawatt method is not techUK's, and it does not offer a different multiplier. And the government's first objection is to a metric that, on Verdant's account, produced the 40,200 figure the government has been citing. If the metric is deeply misleading, that is a problem for the larger number as much as the smaller one.
The scope objection is the strongest of the four. Verdant is counting people inside the building; techUK is counting an economic footprint including supported and indirect roles. Those are different questions, and the 43,500-versus-24,300 split inside techUK's own figure shows techUK is perfectly capable of separating them. What a worker deciding whether to retrain needs is the direct number, and on that measure the two sides are 24,300 against 4,400.
What this means if you are the worker in question
The honest version is unglamorous. Data centre operational roles are real, skilled and reasonably paid, and there are not very many of them per site. That is true on both sides of this argument.
The national projection is not the number that should govern your decision anyway. The number that matters is the operational headcount for the specific site being built near you, and Verdant flags how often that is simply absent: "Six of the largest schemes in the pipeline, together representing close to 900 megawatts, publish a construction figure or an economy-wide claim but no operational" figure at all.
That is a checkable gap. Planning applications are public. If a scheme near you is being justified locally on jobs, the operational headcount is either in the documents or it is not, and a missing number is itself an answer.
One thing this dispute is not about: construction. Verdant is explicit that its figures are "the permanent positions created by the buildout; they are not including the temporary boost to construction work, which is likely to be larger, but limited in duration." Building the thing employs many more people than running it, for a while. Nothing here contradicts that, and nothing here should be read as saying otherwise.
What would settle it
Operational headcount disclosed per scheme, at consent stage, in the planning documents. Verdant had to reconstruct these numbers from planning applications, ministerial statements and press releases precisely because they are not published as a matter of course. If they were, nobody would need a multiplier at all and this entire dispute would be an arithmetic exercise rather than a modelling one.
Until then the defensible summary is narrow. Two organisations agree on the method and disagree on the multiplier by a factor of seven. One has published its sample, its ratios and its own caveat about site size. The other's rebuttal challenges the sample and the scope without offering a replacement figure. And both of their numbers, taken at face value, describe an industry that employs far fewer people per megawatt than the industries it is being compared to.
The Verdant briefing was downloaded and read in full, and the CNBC page was fetched and read, on 10 September 2026. Every quotation above was checked against the document it came from.
Frequently asked questions
How many jobs do UK data centres actually create?
There is no agreed figure. techUK, the industry trade association, claims 43,500 jobs are already supported by the existing UK estate, of which 24,300 are directly in data centres, and projects 40,200 additional direct operational jobs from planned build-out by 2035. Verdant, an economic think tank, rebuilt the estimate from planning applications and public records and puts existing direct employment at around 4,400 and the 2035 figure at around 10,400. Both numbers describe permanent operational roles, not construction work.
What is the jobs-per-megawatt metric?
It is a multiplier that converts a data centre’s electrical capacity into an employment estimate. Verdant states that techUK uses a jobs-per-megawatt formula and that it is "used widely in planning decisions", and Verdant deliberately re-estimates the same multiplier rather than replacing the method. Verdant says techUK’s model assumes 8.5 operational jobs per megawatt; Verdant’s own capacity-weighted average across 20 proposed schemes is 1.2, rising to 1.6 once the size of the schemes actually proposed is taken into account.
Why would bigger data centres create fewer jobs per megawatt?
Because staffing does not scale with capacity. Verdant’s executive summary states that techUK’s figure "does not account for the fact that majority of new UK capacity planned is in “hyperscale” data centres, which (because they operate at such a massive scale) create fewer jobs for each megawatt of computing capacity installed." A larger site needs more power and more racks, but not proportionally more technicians, so a multiplier derived from a mixed older estate will overstate employment when applied to a pipeline dominated by very large sites.
Does techUK dispute the findings?
Yes, on two grounds, both quoted by CNBC on 9 September 2026. techUK said "Verdant offers its own employment estimate using an unreliably small sample size" and that "the report reduces the economic value of the sector to only the number of people physically working inside data centres, not the countless jobs that are supported by data centres and the services they provide." Notably, the statement disputes the sample and the scope. It does not say that the jobs-per-megawatt method is not techUK’s own.
What does the UK government say?
A UK government spokesperson told CNBC that the report "relies on a deeply misleading jobs-per-megawatt metric, makes false comparisons with hospitals and schools, and appears to confuse electricity capacity with actual consumption", and that "it discounts the wider economic value of compute and the national security and sovereignty case for building critical digital infrastructure in Britain." The first objection is worth reading carefully: the metric being called deeply misleading is, per Verdant, the metric the industry’s own projection is built on.
Should I retrain for a data centre job?
Nothing in this dispute says data centre roles are bad jobs. What both sides implicitly agree on is that the permanent operational headcount at a finished facility is small relative to the capital and power involved, and the disagreement is about how small. If you are weighing a move, the useful question is not the national projection but the operational headcount stated in the planning application for the specific site near you. Verdant notes that six of the largest schemes in the UK pipeline, together close to 900 megawatts, publish a construction figure or an economy-wide claim but no operational figure at all.