Our view · the one opinion on this site
Build them, and make them pay their way
Data centres are worth having: Ireland's digital economy runs on them, and Europe has too little computing power. But the deal isn't good enough yet. They used 23.2% of Ireland's metered electricity in 2025 (CSO). They use only about 38–46% of the grid capacity they hold, and new projects are short of the power the rules now require. Our view: build them, and make each one pay its way.
Updated 12 October 2026 · data build 2026-10-10 · how we know · corrections
Source: CSO data-centre electricity 2025 · EirGrid AIRAA 2026–2035 (contracted capacity) · StackPath firm-power gap table · EuroHPC AI Factories · as of 2026-10-10 · Compiled by StackPath
Why have a view at all
Because the argument about data centres runs on slogans, from both sides.
- We fact-checked 12 things people often say about Irish data centres. 9 of them are false or misleading, and they come from both sides: "a fifth of all our energy", "a bottle of water per prompt" and "they pay no tax", but also "876,000 jobs" and "they run on renewables". Every fact check.
- Decisions worth billions (who gets a grid connection, which projects get permission, where the next gigawatt goes, who pays for it) deserve better than that.
- A view gives readers something to test. Ours comes with the evidence for it, the evidence against it, and the case against it in its strongest form.
The case for data centres
What they bring, and the scare stories that don't hold up.
- They pay councils. About €73m a year in commercial rates, on our estimate from the valuation list. Rates by council.
- They employ people. About 3,300 people work in operating data centres, and about 19,500 in building and running them, indirect jobs included. Employment taxes come to about €280m (KPMG, 2024; reported). Jobs.
- They anchor an industry. The platforms and cloud regions that serve Europe from Dublin sit on them. Europe as a whole has too little computing power: data centres used only about 3% of Europe's electricity in 2024 (Ember, estimate), against 23.2% in Ireland.
- Many of the worries are overstated. Water from public supply is a small share nationally, though two sites use most of it (check). Backup diesel runs a few hours a year, not all the time (check). Data centres are about 5% of Ireland's total energy use, not a fifth (check).
The case against the deal as it stands
Where the costs land, and who carries them.
- They take a share of the grid no other country gives them. 23.2% of metered electricity in 2025, against about 3% across Europe. Electricity use.
- They hold far more grid than they use. About 2,400 MVA is connected or contracted (EirGrid), against an average draw of about 875 MW in 2025: roughly 38–46% used (estimate). Meanwhile Dublin has had about one new data-centre grid contract since late 2021. Power and the grid.
- They have reached a grid-safety limit. The grid can lose at most 900 MW of demand at once after a fault, and data-centre load reached that limit at the June 2026 peak (EirGrid/SONI Weekly Operational Constraints Update, week 42 2026). From 27 October 2026, sites that can't stay on through faults are capped.
- Households probably pay some of the cost. On our rerun of a published method, data centres may add up to about €110–130 a year to a household's bill through wholesale prices. It's an upper-end estimate, and the size is uncertain (the full check).
- New projects are short of power. Under the CRU's December 2025 rules, new data centres must bring their own firm power. The 8 pipeline campuses that state their demand (861 MW) describe about 430 MW of it: about 500 MW short (estimate). The gap table.
- Ireland carries the costs, but the new AI capacity goes elsewhere. 0 of the EU's 19 AI Factories are in Ireland. AI Factories.
The test: four checks
What 'paying its way' means, and where Ireland stands on each today. We apply it to the country as a whole. We don't score individual campuses.
| Check | The question | Ireland today |
|---|---|---|
| 1. Uses what it holds | Does it draw the grid capacity it has reserved, or sit on it while others wait? | Not yet: about 38–46% used (estimate) |
| 2. Brings its own power | Does new load bring firm power and new renewables that count, rather than lean on the system? | Not yet: the pipeline is ~500 MW short of firm power (estimate) |
| 3. Covers its grid cost | Does it pay for the connection, stability and price effects it causes? | Unclear: large users' network charges rose sharply to 2024–25 but are set to fall while households' rise; the bill effect is uncertain (check) |
| 4. Leaves something local | Rates, jobs, heat, water: does the place that hosts it gain? | Partly: about €73m a year in rates (estimate), about 3,300 operating jobs (KPMG), one working heat-reuse scheme (Tallaght) |
Each answer links back to the data behind it. "Not yet" means the national picture falls short today; it's not a verdict on any one site.
The case against our view
The strongest arguments we haven't settled, stated as their supporters would.
- "The bill effect is a model, not a measurement." True. Our estimate holds imports and power plants fixed, so treat it as an upper bound. The real figure could be much smaller.
- "Unused capacity is a reserve, not waste." Operators need headroom to grow and to stay resilient, and contracts were signed under the old rules. Using all of it at once would also push the grid harder.
- "Rates are small next to the tax you can't see." Corporation tax from the companies that run these sites isn't published per data centre, so we can't count it either way. It may be much larger than rates.
- "Make them pay, and they'll build in Spain or the Nordics." Europe's new public AI capacity is already going elsewhere: none of the EU's AI Factories is in Ireland. Tougher terms could push private projects the same way. Cheap, clean power elsewhere is a real advantage, and Europe as a whole may be better off building AI capacity where power is plentiful.
- "Bring-your-own-power means more gas." Most firm power for new sites will be gas engines or turbines, at least at first. Making data centres pay their way on the grid could cost more on climate.
What would change our mind: evidence that the bill effect is small, that unused capacity is being released or used, or that new projects are bringing the power the rules require.
How we hold a view and keep the facts straight
- The view lives on this page. Facts on every other page stay neutral, and every company is measured the same way, every hyperscaler included.
- Every number has a source and a date. Estimates are labelled as ours. Mistakes go on the corrections page.
- We don't lobby, take money from the companies we measure, or make asks of officials. About StackPath.
- Disagree? Tell us why at [email protected]. If you're right, we'll change the page and say so.
Follow the evidence as it changes
The Weekly: every Monday, the five things that changed in data centres and power, each with its source.
Questions people ask
- Are data centres good or bad for Ireland?
- Both, on today's terms. They bring rates, jobs and an industry, but they used 23.2% of metered electricity in 2025 (CSO), hold far more grid capacity than they use, and may add to household bills. Our view: build them, and make each one pay its way.
- Is StackPath for or against data centres?
- For building them, on conditions: use the grid capacity you hold, bring your own power, cover the grid costs you cause, and leave something in the place that hosts you. The rest of the site reports facts without taking sides.
- Do you score individual data centres?
- Not at the moment. We apply the four checks to Ireland as a whole. The data for each campus is on its own page.