Britain is rapidly expanding AI data centres, but an analysis says the permanent jobs actually created could be far fewer than industry forecasts.

On Sept. 9, CNBC reported that British think tank Verdant estimated the number of jobs directly supported by data centres in Britain at about 4,400. That is less than one-fifth of the 24,300 cited by industry body techUK. Verdant also projected about 10,400 direct operating jobs including planned facilities, a large gap from techUK’s forecast of an additional 40,200 jobs by 2035.

The difference stems from how the figures were calculated. Verdant analysed staffing data using permitting documents for 20 data centres in Britain and company announcements. It argued that newer large facilities need fewer permanent staff due to automation and economies of scale, lowering jobs per megawatt compared with smaller facilities in the past.

techUK countered that the report’s sample was too small and focused only on direct employment inside data centres. It said the assessment should also cover how data centres support broad economic activity, including supply chains and cloud and AI services. The British government also said it is difficult to judge economic value using jobs per megawatt alone, adding that data centres are also important for national security and digital sovereignty. The government designated data centres as critical national infrastructure in 2024.

Investment in AI infrastructure does not mean jobs do not grow at all. Randstad analysed more than 50 million job postings and found demand has risen 107 percent for robotics technicians, 67 percent for cooling and heating, ventilation and air conditioning engineers, and 51 percent for industrial automation technicians since late 2022. That suggests demand is more pronounced for specialised roles in construction, power, cooling and automation than for permanent data centre staff.

Alphabet, Microsoft, Amazon and Meta also plan capital spending of around $700 billion this year focused on infrastructure such as AI data centres and chips. A debate is expected to intensify over whether the economic value of AI infrastructure, which consumes huge amounts of power and land, can be assessed simply by the scale of permanent employment.