Evidence Series — Scottish Energy Compact — June 2026
The UK Government claimed 3,400 jobs for the Lanarkshire AI Growth Zone. A Freedom of Information request revealed the figure was assembled from temporary construction roles, borrowed estimates, and multipliers of up to 6x. The permanent operational reality, globally, is closer to 150. This is the evidence.
Section 1 — The FOI Evidence
On 29 January 2026, Prime Minister Starmer announced the Lanarkshire AI Growth Zone alongside a specific jobs promise: more than 3,400 jobs over the coming years, including around 800 high-value roles. Technology Secretary Liz Kendall said the benefits of AI would “be felt locally.”
Action to Protect Rural Scotland submitted a Freedom of Information request to the UK Government asking how the 3,400 figure was derived. The response revealed the methodology in full. The trail below reconstructs it step by step.
The UK Government based its Lanarkshire estimate on projected jobs from the Cambois data centre development in Northumberland. Cambois was estimated to directly create 1,200 construction jobs and 400 operational roles.
Before the Lanarkshire announcement, Labour MP Ian Lavery — whose constituency includes Cambois — told the House of Commons that big data centres could expect to employ 1,600 people in the construction phase but only about 150 in the operational phase. The government used 400 operational jobs as its source figure despite this public challenge.
The Cambois figures were scaled upward to reflect the projected larger capacity of the Lanarkshire AI Growth Zone, using employment multipliers of approximately 1.5 to 1.6 applied to both construction and operational roles.
More than 61% of the final 3,400 figure comes from multiplying up for roles assumed to be created in the wider economy — indirect and induced employment. The UK Government applied this multiplier to construction jobs as well as operational roles, despite construction being temporary work lasting 12 to 18 months.
APRS concluded that the government appeared to be taking the actual estimate of jobs, adding temporary construction and indirect employment, and then assuming each of these creates up to six additional jobs in the wider economy — the only way it would be possible to arrive at these figures.
Jobs “over the coming years” including 800 high-value AI and digital roles. Announced by the Prime Minister. Not contractual. Not independently verified. Based on scaled Northumberland figures with 6x multipliers applied.
Permanent operational staff — the figure the local MP cited in Parliament for the Cambois facility that was used as the source data. Global hyperscale facilities consistently employ 30–200 permanent staff regardless of size.
Reconstructed from APRS FOI response. Temporary construction roles last 12–18 months. Indirect/induced roles assume up to 6x multiplier on all other categories. None of these figures are contractual obligations on any developer.
Scottish households are being asked to accept higher electricity bills so that facilities employing fewer permanent staff than a village secondary school can train AI models for clients in California.
Michael Knox — Scotland’s Data Centre Reckoning, June 2026Section 2 — Global Operational Evidence
A modern hyperscale facility is designed for maximum automation and minimum human presence. The industry rule of thumb is 0.5 to 2 permanent employees per megawatt of IT load. The table opposite shows confirmed permanent staff figures from comparable facilities worldwide.
The pattern is consistent across operators, geographies, and facility sizes. A 600MW Scottish facility would, at best, employ 300–600 permanent staff — and many comparable campuses operate at the lower end of 0.5 jobs per MW, meaning around 300 people running a site that consumes the electricity of a medium-sized city.
The roles are overwhelmingly high-skill and specialist: network engineers, power systems technicians, cooling engineers, security-cleared data centre operators. They are routinely filled by internal company transfers or international recruitment rather than local hiring. The genuinely accessible local jobs — security, grounds, cleaning — are few, low-wage, and outsourced to contractors.
Ireland hosts a larger share of Europe’s hyperscale data centre capacity than any other country. Its experience is directly relevant to Scotland. Ireland’s Data Centre Strategic Review 2020 found that despite hosting facilities equivalent to 14% of national electricity consumption, direct data centre employment across all operators combined in Ireland was under 2,500 jobs. That is less than a single medium-sized manufacturing plant — for an installed base representing billions of euros of infrastructure investment.
| Facility | Operator | IT Load | Permanent Staff |
|---|---|---|---|
| Hamina, Finland | ~40MW | 60–80 | |
| Viborg, Denmark | Apple | ~40MW | 50–70 |
| Grange Castle, Dublin | Microsoft | ~250MW | 150–200 |
| Quincy, Washington | Microsoft | ~300MW | ~100 |
| Typical single-AZ site | Amazon | ~100MW | 30–50 |
| Luleå, Sweden | Meta | ~120MW | 100–120 |
| Middenmeer, Netherlands | ~200MW | ~150 | |
| Clondalkin, Dublin | Amazon | ~150MW | ~60–80 |
| Projected Scottish 600MW site | Various | 600MW | 300–600 (best case) |
Sources: Operator corporate filings, planning application supporting documents, facility environmental statements, and publicly available facility profiles. Permanent operational staff only — excludes contracted services, construction, and vendor maintenance roles.
THE MICROSOFT QUINCY COMPARISON
Microsoft’s Quincy, Washington campus is one of the largest hyperscale facilities in the world at approximately 300MW — comparable in significance to what is proposed for Scotland. It employs around 100 permanent staff, of whom approximately 30 are in IT-related roles. The rest are facilities, security, and management. This is the realistic employment model for a Scottish hyperscale site.
Source: Microsoft facility disclosures; Grant County PUD employment data
Section 3 — Investment Efficiency
The jobs argument for data centres does not just fail on the headline figures. It fails on the fundamental relationship between capital investment and employment creation. Longitudinal analysis of Virginia — the most data-centre-dense jurisdiction in the world — provides the clearest evidence.
APRS analysis of Virginia investment data over the past decade found that every job created from data centre investment cost $33 million in capital. Across the full Virginia data centre industry, $71 billion of investment created fewer permanent jobs than $34 billion of manufacturing investment over the same period — and manufacturing created 56 times more jobs per dollar invested.
Jobs created per £100 million invested. Source: APRS analysis of Virginia longitudinal investment data, March 2026.
THE VIRGINIA LESSON
Northern Virginia data centre investment over the past decade: $71 billion. Manufacturing investment over the same period: $34 billion. Manufacturing generated 56 times more jobs per dollar. Virginia is now actively debating whether data centre tax incentives represent value for public money — a debate Scotland has not yet begun, despite approving a pipeline twice the size of Virginia’s annual build rate.
Source: APRS Jobs and Data Centres report, March 2026; Virginia Joint Legislative Audit and Review Commission
Proponents routinely cite economic multipliers claiming each data centre job supports three to five others in the wider economy. The Virginia longitudinal data directly tests this claim — and finds it does not hold for hyperscale facilities.
The reasons are structural. Most equipment — servers, network switches, cooling plant, UPS systems — is manufactured outside Scotland and replaced on vendor contracts with multinational firms (Dell, Cisco, Vertiv, Schneider Electric). The supply chain for a hyperscale facility does not run through Scottish manufacturers or suppliers. A campus of 200 highly paid engineers may spend locally, but if they are international transfers or remit a significant share of income abroad, the net local injection is modest.
Ireland’s experience is the definitive test. Despite two decades of hyperscale development, the multiplier effect has not generated a measurable technology employment boom in the host communities. The data centre industry in Ireland employs fewer people than a single large manufacturing facility — while consuming 14% of national electricity.
KEY SOURCES FOR THIS SECTION
APRS — Jobs and Data Centres report, March 2026
APRS — FOI response on Lanarkshire AI Growth Zone employment, April 2026
Hansard — Ian Lavery MP, Cambois data centre employment
BBC — Cambois data centre jobs reporting
Ireland Data Centre Strategic Review 2020
Microsoft — Quincy, WA facility disclosures
Grant County PUD — Washington State employment data
Virginia JLARC — Data centre tax incentive review
Uptime Institute — Global Data Centre Survey 2024
BM Magazine — Lanarkshire AI Growth Zone announcement, Jan 2026
Stop Climate Chaos Scotland — Data centre briefing, May 2026
Section 4 — What Must Change
The employment case for hyperscale data centres is not weak — it is, on the evidence, the primary public justification offered by both the UK Government and developers for subsidised electricity, preferential planning treatment, and socialised infrastructure costs. And it is built on figures that do not survive scrutiny.
The Scottish Energy Compact does not argue that Scotland should reject data centre investment. It argues that if employment is the justification for public subsidy and consent, then employment obligations must be contractual — not press release promises with no legal force whatsoever.
Binding employment target. Every application above 200MW must include a legally binding Section 75 employment obligation specifying minimum permanent operational headcount within three years of commissioning, with financial penalties for non-compliance.
Local hiring obligations. A minimum 40% of permanent operational roles must be offered to candidates resident within 25 miles of the facility at time of application, with a community skills development fund contribution of £5,000 per role where this threshold is not met.
Construction labour standards. All construction contracts must specify Scottish and UK labour for a minimum 60% of site hours, with HMRC-registered employment (no bogus self-employment), Living Wage compliance, and published workforce data quarterly.
No phantom multipliers. No jobs figure cited in any planning application, environmental statement, or public announcement may include indirect or induced employment in the headline number without clear, separate disclosure. The permanent direct operational headcount must be stated prominently and separately.
Pending the full Compact, one change can be made immediately:
PLANNING GUIDANCE — NOW
The Scottish Government should issue planning guidance to all local authorities requiring that any jobs figure cited in a data centre planning application must separately state: (1) permanent direct operational headcount; (2) temporary construction headcount with duration; (3) indirect/induced estimates clearly labelled as non-contractual projections. This requires no new legislation — it is a planning guidance note.
THE BROADER POINT
The jobs argument is the weakest element of the developer’s pitch, and therefore the strongest opening for demanding real returns. If a developer cannot commit to contractual employment obligations, they cannot claim employment as a justification for public subsidy. That is not an unreasonable position. It is the basic standard of commercial accountability.