Poorest households spend twice the national share of income on water — the data centre boom will push bills higher

The cost of building the power and water systems the data centre boom needs does not vanish. It lands on household bills. New research from The Firewalkers finds that this cost falls hardest on the poorest and most rural families. They already spend a far larger share of their income on energy and water — and they are the least likely to benefit from the buildout next door.

The research draws together official forecasts across the UK, Ireland and the European Union to ask a single question that planning systems rarely answer out loud: when the grid and the reservoirs have to expand to serve hundreds of new data centres, who actually pays — and can they afford it?

Why the cost is regressive by design

Utility bills are largely flat-rate for small consumers. A poorer family does not get a cheaper unit of electricity than a wealthy one — so when wholesale prices rise to fund new generation and grid upgrades, the same percentage increase consumes a much larger slice of a low-income budget. A 1% rise is a rounding error for an affluent household and a real choice between heating and food for a struggling one.

The research sets out how stark that gap already is. Domestic energy costs make up about 6% of total spending for the average UK household — but around 10% for the lowest-income tenth, according to official spending data. Water is the same shape: typical bills are around 1.6% of income, but among the poorest third they reach 3.7%. The recognised thresholds for water poverty are 3% and, for severe hardship, 5%. The poorest third are already at the edge of the first.

That is the situation before the data centre boom adds new demand. The expansion does not create these inequalities. It pushes on the exact point where they are already most painful.

What the buildout does to the bill

The research identifies two ways data centre growth pushes prices up. The first is new generation: meeting sudden peaks and low-wind periods increasingly relies on gas plants, whose costs flow through to wholesale prices. The second is infrastructure — the substations, transmission lines and reservoirs needed to connect and supply the new load. Those costs typically feed into the tariffs paid by all customers.

The clearest warning comes from the United States, where the pattern is further advanced. Virginia hosts the world's densest data centre cluster. A 2024 study by the state's own Joint Legislative Audit and Review Commission found that if data centre growth continues unconstrained, Virginia's electricity use could rise 183% by 2040 — against just 15% if there were no new data centre demand. Almost all of the projected growth is driven by data centres, not by population or electric vehicles.

That demand has a cost, and households are already feeling it. Residential electricity bills in Virginia have risen by close to 30% since 2021, faster than general inflation, according to analysis of the utilities' own regulatory filings. In response, Virginia's regulator approved a new rule in early 2026 requiring large data centre users to pay a minimum share of the distribution, transmission and generation capacity they contract — explicitly to protect other ratepayers from absorbing those costs.

A note of caution on a widely-shared figure: claims that bills near data centres have risen "267%" refer to wholesale market prices, which are only one component of a household bill — not to the bills households actually pay. The Firewalkers does not use that number. The residential reality — roughly 30% in Virginia since 2021 — is serious enough without overstatement.

The UK is not on US-style contracts, but the underlying mechanism is the same. Ofgem has warned that around 140 proposed data centre projects could require some 50 GW of new power capacity — about 5 GW more than today's entire national peak of 45 GW — and that connecting these loads requires substantial investment in new substations and lines. Build that infrastructure on the assumption of demand that may not all arrive, and it is billpayers who cover the gap.

One number is conspicuously missing from the UK debate: how much this adds to the average household bill. No UK body has yet published a clear, public estimate of the per-household cost of the grid expansion the data centre queue implies. That absence is itself a finding. Decisions worth tens of billions are being taken without the one figure an ordinary billpayer would most want to see.

The subsidy question — who is being protected, and from what

Data centre developments often arrive with public support. The UK's AI Growth Zones offer qualifying sites large discounts on wholesale power and faster grid connections. The research notes a single 500 MW centre in Scotland could save around £24 per megawatt-hour — up to roughly £80 million a year — if it helps stabilise the grid.

The official position is that this support is self-financing and costs other billpayers nothing, because the sites sit near surplus renewables. That may hold. But it is a forecast, not a guarantee. If the forecast overshoots, the risk has been socialised: consumers cover the grid build through their rates while a private operator keeps the discount. The pattern is visible in Northern Virginia, where data centres generate substantial local tax revenue — but where serving their demand requires a far larger programme of grid investment, much of it recovered from residents and consumers rather than the operators.

The rural edge of the problem

Rural and low-income communities sit at the sharpest point. Many rural households depend on scarcer water supplies and on agriculture, which competes for the same water in a drought. In Ireland, urban households account for a larger share of metered electricity demand than rural ones — reflecting sparser, less efficient rural living and higher per-capita costs. When prices rise, these are the households with the least slack to absorb them.

The research is blunt about the imbalance. Data centres mostly bring short-term construction jobs and tax revenue, not lasting local employment. The communities asked to carry the higher bills, the water stress, and the grid strain are frequently the ones least likely to see any ongoing benefit in return.

What can be done — and it is not "stop building"

The research sets out practical policy options. The first is cost internalisation: developers should fund the grid and water upgrades their projects require, through deposits, capacity auctions, or build-and-transfer programmes, rather than passing them to billpayers. Ireland already requires new connections to include on-site generation or storage. The UK's emerging Growth Zone model moves in this direction by letting operators build their own transmission lines.

The second is protecting vulnerable users directly. Any residual price rise driven by data centres should be cushioned by enhanced social tariffs or rebates for low-income households. A small per-kilowatt-hour levy charged only to data centre sites could fund energy-bill assistance for the poorest households — turning a source of pressure into a source of protection.

The third is transparency. Operators should report their energy and water use publicly, so communities can see the real impact rather than relying on assurances. None of this stops a single useful data centre being built. It changes who pays for it.