Data centres are the industrial engines of the AI revolution. These vast buildings, housing the computers used by large language models, are springing up all over the developed world. Britain has an estimated 560 data centres — some built several years ago to support cloud computing, and some appearing more recently during the AI revolution. The race to build infrastructure fit for the 21st century, however, is exposing weaknesses in Britain’s industrial base. One bottleneck is the inefficiency of the electricity grid. Another is a tangle of regulation that slows planning and deters investment. These issues must be addressed for Britain to live up to its potential.

Industrial energy prices in Britain are cripplingly high — four times those in America. Given that modern AI systems require a large amount of computational power, the cost of energy creates an acute barrier to investment in data centres. The problem is that in Britain the wholesale energy price is set nationally and tends to allow comparatively costly gas-powered plants to act as a “price setter”. That pushes up prices, as well as depriving the market of signals: it means industry cannot move to areas where an abundance of energy ought to make production much cheaper. Many industry analysts argue that the grid should move to a system, like America’s, using variable “locational prices”. Octopus Energy calculates this switch could save consumers as much as £55 billion by 2050. It could drive down Scotland’s energy prices to some of the cheapest in Europe. It would also allow energy-hungry industries to be located closer to where that energy is generated, avoiding inefficient energy transmission.

The electricity grid is also too weak. Ofgem, the energy regulator, proposes to charge developers an upfront fee of as much as £712,500 per megawatt for projects planning to connect to the grid.

That is a sensible interim step. At present it can take months or years to connect to the grid — a huge liability in an industry where speed is of the essence. Such frictions are thought to make the lifetime value of a 100-megawatt data centre almost a fifth lower in Britain than in America.

By forcing developers to pay up front, Ofgem is hoping to weed out unserious and speculative planners from among the 315 data centre proposals in the long queue for approval. This measure is, however, ultimately a concession to the underlying weakness of the grid. Britain is unable to meet the demand of the investors courting it. It is estimated that the total electricity capacity required by projects planning to join the grid has recently increased to as much as 125 gigawatts: almost three times the peak electricity demand last year. Britain’s underlying shortage of energy capacity is another bottleneck to be addressed.

As more data centres appear across Britain, other forms of opposition will rear their heads. In America, nimbyist concerns about the pollution and unsightliness caused by data centres have fuelled a political backlash against the technology. The Scottish National Party has recently called for a complete moratorium on the construction of data centres north of the border.

Last year, the London Assembly’s planning and regeneration committee alleged that data centre construction might be diverting limited grid connectivity and resources away from housebuilding.

It is clear, however, that London will need data centres close at hand, not least to serve the capital’s finance sector.

These objections offer an intimation of the political football that will be made of data centres in years to come. Ill-motivated blockers must not be allowed to stop Britain from building infrastructure where there is a sound business case for doing so. Nor should they distract policymakers from addressing the UK’s deep structural weaknesses, now so glaringly exposed in the age of AI.