The rapid scaling of artificial intelligence infrastructure in the UK is hitting an unexpected but critical obstacle — a shortage of water resources. Industry analysts have concluded that government forecasts for water consumption are "fatally flawed" and completely ignore the demands of data centers.

The problem lies in the physics of how modern data centers operate. The enormous computing power required for training and running AI models generates a colossal amount of heat. Cooling towers and liquid cooling systems, which consume water directly, are used to dissipate this heat. Additionally, indirect resource consumption grows alongside the data centers' electricity needs, as power generation also requires water.

Ambitions vs. Reality

An analytical document presented to parliamentarians reveals a serious gap between the government's ambitious plans for AI development and actual resource planning. In key strategic documents — from policies on creating AI growth zones to the national water program — there is not a single mention of data center needs. This makes current plans to triple computing capacity by 2030 practically unfeasible without a radical overhaul of approaches.

The situation is particularly acute in the Slough region, where the majority of British data centers are concentrated. The local water operator, Affinity Water, is already recording requests from data centers at the level of 3 million liters of water per day — comparable to the peak consumption of a small town with a population of around 3,500 households. Meanwhile, the region is already suffering from a shortage, as evidenced by an existing ban on hose watering.

The situation is exacerbated by current regulations requiring data centers to use only drinking water. According to estimates, daily consumption for these purposes already amounts to about 6.6 million liters. If the government fulfills its promises to increase capacity, this figure could jump to 19.8 million liters per day.

"The only solution to ensure new supplies is either to increase water extraction from the environment, which would harm it, or to build new infrastructure facilities. However, planning and constructing large reservoirs can take up to 15 years," notes the Water UK document.

Notably, similar problems have already forced New York authorities to impose a moratorium on the construction of large data centers, while in Texas, data center operators are even trying to bypass public environmental oversight. The UK risks falling into a similar trap, where technological progress hits a wall of basic resources.

Expert opinion: The water issue could become a more serious constraint for the AI industry than a shortage of chips or electricity. Investors and data center operators urgently need to rethink deployment strategies and implement closed-loop water supply systems; otherwise, ambitious plans for scaling AI risk remaining on paper.