So far, fears have focused on the pressure data centres exert on local energy costs. This is why, in August, Pennsylvania Governor Josh Shapiro imposed steep requirements on developers, and why many U.S. municipalities have paused new projects. In the Netherlands, the halting a Meta campus in Zeewolde in 2022 was a landmark moment.
This backlash isn't all populist Luddism. Research by Yale College's Ryan Wang estimates that adding 100 megawatts of operational data-center capacity in a U.S. county raises residential fixed electricity charges by 0.7% in regulated markets. While local GDP rises 0.78%, the growth is driven almost entirely by data centers themselves, and a lot of the gain is the one-off construction bump, whereas unemployment rises by 0.7% and manufacturing wages fall by 0.5% — likely because companies try to recoup lower earnings from higher electricity bills. Translate this to mega projects such as the 4.1 gigawatt Stargate build-out in Taylor and Shackelford Counties in Texas, and Wang's model spits out a devastating 8.3% fall in local inflation-adjusted wages.
Still, the better question is whether eventually, when bottlenecks are overcome, data centres create permanent, well-paid tech and telecommunications jobs. A Brookings analysis found that facilities built by specialised landlords like Equinix don't, because they tend to serve remote users, but campuses built by hyperscalers do. Indeed, Northern Virginia and Dublin have both used data centres as a stepping stone to become major tech hubs, because officials pursued a coherent economic strategy, investing in energy, infrastructure, and expanded university programmes.
As Jane Flegal of the Searchlight Institute argues, AI mania could thus create the political opening needed to fund long-needed upgrades to energy transmission and loosen planning rules. The 2025 U.S. deal with Westinghouse owners Brookfield Asset Management and Cameco to build at least $80 billion of new nuclear reactors is one example.
The missing piece is to add housing shortages to the equation. Ireland is a cautionary tale: attracting tech investment proved easier than housing the workers it brought. In the near term, data centres, infrastructure and housing compete for the same builders. In the long run, though, governments could use hyperscaler investment to support larger housing targets. In the UK, for instance, Andy Burnham's government recently pledged to build 70,000 social homes over a decade, which is rather lacklustre. If officials tied major data-centre developments such as Blackstone's 720 megawatt campus in Blyth — a post-industrial laggard in north-east England — to larger residential construction plans and training schemes, perhaps such ambitions could be raised substantially.
To find precedents, look at the big postwar reconstruction in Europe under the Marshall Plan, which simultaneously rebuilt homes, factories and transport networks. Even in the U.S., higher residential building started in the 1950s alongside mammoth projects like the Interstate Highway System. This was possible because a long pipeline of work kept construction employment sustainably at around 5.5% of the total non-farm workforce or above, even during recessions. Since the 1980s, the average has been 4.8%.
Officials are now caught between two extremes: reflexively blocking data centres or embracing them with little regard for the strain they put on resources. The winners will be those who fall for neither.
QQQ•
