Three ways data centres could reach a household bill
Data centres used 7.7 TWh in 2025, 23.2% of metered electricity (CSO). They could affect what households pay in three ways that can be checked with public data, and some that cannot:
- Wholesale prices. Extra demand means dearer plant is needed more often.
- The PSO levy, which funds renewable support and is shared out by a formula.
- Network charges, which pay for the grid and are split between customer groups by the CRU.
- Not testable: who pays for grid reinforcement that data centres triggered, capacity payments and constraint costs.
Where Ireland's electricity growth went
Since 2015, metered electricity use grew by 8,388 GWh, and data centres account for 77% of that growth.
Each bar is all electricity measured by meters in Ireland in one year: red is data centres, grey is homes, businesses and industry combined. Data centres went from 1,240 GWh (5%) in 2015 to 7,662 GWh (23.2%) in 2025, while everyone else added 1,966 GWh. The share of growth is our arithmetic on the CSO series. The SEAI, measuring all electricity demand rather than metered use, puts data centres at 88.2% of demand growth from 2015 to 2024.
Sources: CSO, Data Centres Metered Electricity Consumption (MEC02), 7 Jul 2026 · SEAI, Energy in Ireland 2025, Dec 2025
Wholesale prices: the €360 estimate, rerun
The May 2026 study for Friends of the Earth Ireland, by Seán Fearon, estimated that data centres added about €360 to the average household's electricity costs over 2015–2023, €263 of it in 2021–23. Its method asks how much the price rises with "residual demand" (demand minus wind) and attributes the data-centre share of that demand.
We reran the same method on independent data (Ember hourly prices and EirGrid demand and wind for 2021–23; SEMOpx and EirGrid for the last year). It gives:
So the number reproduces in size. It is, though, a short-run model: it holds imports, the plant fleet and bidding fixed and treats the whole link between demand and price as caused by load. That makes it an upper-end estimate of the wholesale effect, not a measured bill impact. The study itself hedges its result ("appears to be"), and the method leaves out channels that could add to the cost, such as capacity payments and constraint costs. Suppliers also hedge, so wholesale changes reach bills late and only in part. The study's projections for 2025–2034 depend on demand and fuel scenarios and cannot be tested.
How the wholesale price is set, and what a flat load would pay
The last, dearest unit needed sets the price each half-hour. A data centre draws the same power all day, so at spot prices it would pay the plain time average, €136.9/MWh over the last year; wind earned 14.7% less, €116.7.
In the day-ahead market, offers are ranked from cheapest to dearest and the price in each half-hour is set where that stack meets demand, so on still evenings older gas plant sets a high price and on windy nights the price falls. Across the last 12 months (2025-10-04 to 2026-10-09) the Irish price averaged €136.9/MWh, with 13.1% of half-hours above €200. A flat load like a data centre buying at spot prices would pay that time average, 1.3% below what the national demand profile paid (€138.7). Wind earned €116.7 because it produces most when prices are low; a flat load still has to buy power when the wind is not blowing, so a wind contract alone does not cover it. Most large users hedge, so this is a spot benchmark, not what any site pays.
EstimatePanel 1 is a schematic of how a pay-as-clear auction works (general description, unchecked against the SEM rulebook; "often" is not measured here). Energy prices only: bills add network, capacity, PSO, taxes and supplier margin.
Sources: SEMOpx day-ahead market results (ROI half-hourly index price), 2026-10-09 · EirGrid Smart Grid Dashboard (demand and wind weights), 2026-10-10
The PSO levy: households pay more than their energy share
The CRU shares out the PSO levy by each customer group's contribution to peak demand. Data centres run flat around the clock, so they add little to the peak. In 2025/26 households carried 34.2% of the levy while using 27.5% of metered electricity (CSO MEC05). Medium and large users, including data centres, carried 55.7% while using about 62%.
The amounts are small. The 2025/26 levy was €162m (revised to €125m from December 2025), €2.01 a month per household before the cut. Sharing it by energy instead would take roughly €5 a year off each household (our estimate). The levy shrinks when wholesale prices are high, because renewable support then pays out less.
Network charges: rose faster for big users, now reversing
The Irish Examiner reported that large energy users' network charges rose 175%, against 45% for households. The CRU's PR6 impact tables reproduce the +175% exactly for its "extra-large energy user" archetype (190 GWh a year, data-centre sized): €2.45m in 2020-21 to €6.76m in 2024-25. The household archetype went from €293 to €384, +31%. We could not reproduce the +45%; it may use 2025-26 tariffs (unchecked).
| CRU archetype | 2020-21 → 2024-25 | 2024-25 → 2029-30 |
|---|---|---|
| Extra-large energy user | +175% | −20% |
| Household | +31% | +15% |
CRU2025198, PR6 Final Determination, Impact Analysis Note (Tables 1, 16, 17). Archetypes are illustrative customers, not averages.
From 2024-25 to 2029-30 the CRU projects the reverse: extra-large users down 20%, households up 15%, as transmission charges (paid mostly by big users) fall and distribution charges rise. Network costs were 31.8% of a typical household bill in the CRU's 2020 factsheet.
A Trinity College study reported by the Irish Times argues that connection rules leave households paying for reinforcement data centres caused. No public data splits reinforcement cost by the customer who caused it, so that claim can't be tested. One supporting fact checks out: the CRU said new capacity at Castlebaggot 220 kV, planned for housing and west Dublin growth, was "almost entirely used by data centres".
The claims, checked
| Claim | Source | Our rating |
|---|---|---|
| Large energy users' network tariffs rose much faster than households': +175% vs +45%, 2020-2026. | Irish Examiner, Sep 2026 | mostly accurate |
| CRU rules socialise data-centre grid costs onto households; connection fees don't cover reinforcement (TCD, 'By Design, Not by Default'). | Irish Times, Sep 2026 | can't be tested with public data |
| Data-centre demand pushed up wholesale prices: ~EUR 360 per household over 2015-2023. | RTÉ, May 2026 | mostly accurate |
| Data centres added ~EUR 360 to the average household's electricity costs over 2015-2023 (EUR 263 in 2021-23) and will add EUR 295-644 over 2025-2034. | Friends of the Earth Ireland, May 2026 | mostly accurate |
| Data-centre demand raises power-system costs by 6% (carbon policy) to 9-15% (renewable targets). | ESRI, Apr 2020 | can't be tested with public data |
| The PSO levy is allocated by contribution to peak demand, so households and SMEs pick up part of the extra renewable subsidy that data-centre demand creates. | ESRI, Jun 2019 | mostly accurate |
Why it matters
Our reading. Our verdict on "data centres are making my bill higher" is partly true (size uncertain). The wholesale effect is real in direction but modelled, not measured, and the best available method gives an upper-end figure. The levy split leans slightly against households; network charges leaned against big users until now and are set to lean the other way. A measured answer would need the CRU or suppliers to publish how wholesale costs and network investment are passed through to each customer group.