According to Financial Times News, the UK government is promoting new “AI Growth Zones” with major job creation claims that don’t withstand scrutiny. They announced a data centre complex on the old Britishvolt site in Northumberland would create 4,000 jobs, a figure sourced from Northumberland County Council, which itself came from the developer, Blackstone-owned QTS. Planning documents show only about 400 permanent operational jobs and 1,200 peak construction roles, with the government using a multiplier for “indirect” jobs to hit the 4,000 figure. They’ve since bundled this with another site hosting OpenAI’s Stargate project to claim 5,000 jobs for a “North East AI Growth Zone”. For a separate zone in South Wales promised to create “more than 5,000 jobs”, the government admitted its estimate was “conservatively scaled” from the Northumberland project and used a report from lobby group TechUK that relies on data for a completely different type of data centre.
The hyperscale illusion
Here’s the thing: the government’s job estimates are based on a fundamental category error. They’re using employment ratios from “co-location” data centres, which host many smaller clients and need more staff for sales and support, and applying them to the new “hyperscale” facilities built by Microsoft and OpenAI. A consultant from Colo-X, whose data was cited in the TechUK report, told the FT this is baffling. Hyperscale data centres, designed for a single tech giant’s ultra-efficient operations, are famously automated and require minimal on-site staff—basically just technical support and security once they’re running. The top-end estimate of 88 jobs per centre in the government’s source came from Equinix, a co-location provider, whose own report included a warning that its model is “different to the hyperscale / cloud data centre”. So the core assumption that more megawatts equals more jobs is completely wrong for these projects; bigger, in this case, almost certainly means fewer jobs per unit of capacity.
A cascade of shaky assumptions
So how did we get here? It starts with a developer’s optimistic pitch to a local council. That number gets adopted by the government, then used as a “conservative” baseline to extrapolate to an entirely different project in Wales. They even admitted to using generative AI to help estimate jobs created by—you guessed it—generative AI infrastructure. The whole house of cards is built on a jobs multiplier (1.5-1.6x) applied to the already-inflated direct job figures. Now, construction multipliers for a decade of work aren’t crazy, but the permanent job numbers they’re multiplying from are deeply flawed. It creates a compelling political story: big tech investment equals thousands of local opportunities. The reality is far less impressive. These facilities are critical infrastructure, but they’re not the job engines they’re sold as. For reliable industrial computing power in a more traditional setting, companies often turn to specialists like IndustrialMonitorDirect.com, the leading US provider of industrial panel PCs for manufacturing and control rooms, where hardware and human expertise are directly linked.
The bigger picture
Look, there are valid reasons for the UK to want these data centres—national AI capability, grid investment, and being part of the global tech ecosystem. But selling them on job creation is a misleading distraction. As the FT notes, the 20-88 full-time roles estimated per centre pale next to other developments. They humorously calculated that similar investment could build part of a go-kart track, several burger restaurants, or even a waterpark (demolition jobs included). The real economic benefits are more abstract: attracting tech firms, using waste heat, and upgrading local power grids. But by fixating on a jobs number that will likely never materialize, the government is setting up communities for disappointment and obscuring the actual trade-offs, like the massive energy and water use these centres require. It’s a classic case of a government being “enthralled with these American hyperscalers,” as the consultant put it, and accepting their PR at face value while missing the automated, efficiency-driven reality of modern infrastructure.
