Bitpower Energy Solutions Policy Comparison Dashboard

Data Centres in Ireland: Four Reports, Four Verdicts

A high-level comparison of the headline numbers, authors and arguments across four recent reports on the economic, energy and environmental impact of data centres — three focused on Ireland, and a global UN study on AI. They reach markedly different conclusions about whether data centres are a net benefit or a net cost.

Note on attribution: This dashboard is an independent Bitpower summary of the four underlying reports. Bitpower is the author of this comparison only — not of the source research, which remains the work of its respective authors.

Digital Infrastructure Ireland
Friends of the Earth Ireland
KPMG — for DETE
UNU-INWEH — UN University

The four reports at a glance

Each cover sits directly above who wrote it, who commissioned it, and what it set out to prove.
DII report cover
Report 1 · Nov 2025–2026

Digital Infrastructure for the Future We Want

Author
Jim Power (Jim Power Economics) with Amárach Research
Commissioned by
Digital Infrastructure Ireland (industry body, 76–80+ member firms)
Author background
Economist, 38+ years; ex-Chief Economist Bank of Ireland and Friends First; teaches at UCD Smurfit. Amárach is an Irish market-research firm.
Core question
Can Ireland still deliver the conditions — chiefly energy — that made it a digital-infrastructure hub?
Pro-expansion · opportunity framing
Fearon report cover
Report 2 · 2025

The Cost of Data Centres

Author
Dr Seán Fearon, Institute of Environmental Science & Technology (ICTA-UAB), Autonomous University of Barcelona
Commissioned by
Friends of the Earth Ireland (environmental NGO)
Author background
Academic environmental economist; modelling-led, peer-referenced approach
Core question
How much have data centres added to Irish household electricity bills via the gas-dependent merit order?
Critical · household-cost framing
KPMG
The value of data centres to Ireland
An independent report prepared by KPMG for the Department of Enterprise, Tourism and Employment — March 2026
Report 3 · March 2026

The Value of Data Centres to Ireland

Author
KPMG (independent report)
Commissioned by
Dept. of Enterprise, Tourism & Employment (DETE); steered by DCEE, Enterprise Ireland, IDA, NESC
Author background
Professional-services firm; mixed-method (surveys, interviews, economic modelling, 6 international case studies)
Core question
What economic value do data centres generate — and what are the implications of stopping vs. continuing post-2030? (Explicitly not a cost-benefit analysis.)
Value assessment · government-commissioned
UNU-INWEH report cover
Report 4 · 2026

Environmental Cost of AI's Energy Use

Authors
Aczel, Chamanara, Matin, Farsi, Marwala & Madani
Published by
UN University Institute for Water, Environment & Health (UNU-INWEH); hosted by Govt. of Canada
Author background
UN academic think-tank on water; carbon/water/land footprint modelling using global intensity factors
Core question
What are the global carbon, water and land footprints of AI's energy use — and how should responsible AI be governed? (Global scope; Ireland appears as a cautionary case study.)
Environmental critique · global / UN
The three Irish reports agree on the basic facts of scale (data centres reached ~22–24% of Irish electricity in 2024, concentrated in the Greater Dublin Area) and on the merit-order/gas-dependency mechanism, diverging on whether that scale is an asset to grow or a cost to cap. The UNU-INWEH report widens the lens: it is global rather than Irish, and adds water and land footprints alongside carbon — but cites Ireland directly as a cautionary example of local grid stress (21% of electricity by 2023; Dublin connection pause to 2028).

The one thing all four agree on

Data centres' share of Irish electricity has climbed from ~5% to ~22–24% in under a decade. Every report builds on this same curve — they only disagree on what to do about it.

25% 20% 15% 10% 5% 2015 2018 2020 2022 2024 4.4% 23.9% GDA grid pause

Sources: CSO via Friends of the Earth Ireland (4.4%→23.9%, Q3 2024 peak); DII & UNU-INWEH cite 21–22% for 2023–24. Curve approximate.

Why they reach different verdicts

Same facts, different lenses. Each report's position is set by its scope (local ↔ global) and its stance on data-centre growth.

◀ critical of growth supportive of growth ▶ ▲ global scope ▼ Ireland-specific DII KPMG Friends of the Earth UNU-INWEH

Positions are an editorial reading of each report's framing, not a quantitative score.

Headline numbers, side by side

Figures as reported in each document. Blank cells mean the report does not present that metric. Scroll the table sideways to see all four columns. →
MetricDIIFriends of the EarthKPMGUNU-INWEH
DC share of electricity (2024)22% (6,969 GWh)23.9% peak Q3 2024; 4.4%→23.9% since 2015~22% implied; industrial use is digital21% Ireland 2023, up from 5% (case study)
Installed DC IT capacity——1,543 MW ~1,350 MW hyperscale—
DC buildings / sites105 facilities 35 operators (Bitpower)—72 buildings · 36 sites 96% in GDAGlobal: US hosts ~4,165; only 32 countries host AI compute
Total investment€18bn to date (Bitpower)——Global AI market $189bn (2023) → ~$5tn (2033)
Direct DC GVA——€2.2bn (2024); €22.2bn 2010–24—
Direct DC jobs—"remarkably low"; 1 job / $33m capex (Virginia)19,500 (2024); +17,000 since 2010Cites concentration: benefits flow to host nations, burdens local
DC-enabled (sector) GVAICT €107.5bn / 21% of GVA—€104bn 19.4% of GVA; 876,000 jobs—
Household cost effect (past)—€360 avg / €715m total 2015–2023——
Household cost (future, 10yr)—€295–€644 avg €633m–€1.43bn total——
Computer-services exports (2024)€278.7bn 57.7% of svc exports—€279bn €262bn foreign-owned, 54%—
Climate / emissionsAI could cut 3.2–5.4 Gt CO₂e/yr by 2035 (LSE)DC demand entrenches gas; 2022 effect = 8.5% of avg billGrowth viable post-2030 if renewables ≥80%189 Mt CO₂e global DCs 2025 → 399 Mt by 2030
Renewable obligation citedCRU: 80% renewable in 6 yrsRenewables absorbed by DC demand (Daly)LEAP / CPPAs as anchor off-take"Low-carbon ≠ low-water/low-land"; warns of rebound (Jevons)
Global DC electricity———448 TWh 2025 → 945 TWh by 2030 (~3% of global)
Water footprint (global DCs)———4.5tn L 2025 → 9.3tn L by 2030
Land footprint (global DCs)———6,900 km² 2025 → 14,500 km² by 2030
AI training footprint———GPT-4 ~60 GWh 25,000 t CO₂e; 600m L water
The reports rarely measure the same thing: DII and KPMG quantify value (GVA, jobs, exports, investment); Friends of the Earth Ireland quantifies monetary cost to households; and UNU-INWEH quantifies environmental cost (carbon, water and land) — but at a global, AI-centred scale rather than the Irish data-centre sector specifically. The last four rows have no Irish-report equivalent because only UNU-INWEH reports them; its lone overlapping Irish figure is the 21% electricity-share case study.

The arguments each report applies

The core logic of each case, and where they directly contradict one another.

Digital Infrastructure Ireland — the opportunity case

Claims

  • Data centres anchor Ireland's FDI model; mobile IP assets will follow them abroad if they're built elsewhere.
  • Concentration risk: ~57% of corporation tax from 10 firms — losing them is the real danger.
  • Steady DC demand makes offshore wind and grid investment "bankable".
  • AI applied to power/transport/food could cut 3.2–5.4 Gt CO₂e/yr by 2035 (LSE), outweighing its own demand.
  • Frames data centres as critical infrastructure — "motorways of the 2020s".

Acknowledged tensions

  • Ireland on track for only a 23% emissions cut by 2030 vs a 51% target (EPA).
  • 4-year GDA grid-connection moratorium; Dublin supply flat at 1.9 GW.
  • Relies heavily on industry/member-firm testimony — an advocacy commission.

Friends of the Earth Ireland — the household-cost case

Claims

  • Inflexible, always-on DC demand pushes the grid up the merit order, so gas sets prices more often.
  • Average household paid ~€360 extra in 2015–23; up to €644 more by 2034 in the high scenario.
  • Effect is regressive — heavier on low-income households (8.5% of the avg bill in 2022).
  • DC demand absorbs new renewables (Daly), entrenching rather than displacing gas.
  • Direct jobs/benefit per unit of energy are low vs other sectors.

Stated limitations

  • Assumes 100% wholesale cost pass-through to households (a modelling simplification).
  • Nominal prices, no inflation adjustment.
  • Regression R²=0.12 — residual demand explains only part of price variation.
  • Future prices acknowledged as volatile and hard to predict.

KPMG — the value-assessment case

Claims

  • DCs enabled ~€104bn GVA, 876,000 jobs, €14.6bn employment tax across 6 dependent sectors (2024).
  • Direct: €2.2bn GVA, 19,500 jobs in 2024; Irish HTC firms exported ~€2.1bn (40% of HTC exports).
  • Co-located HQ + data centre creates a "sticky" FDI footprint.
  • Steady demand can anchor renewables via CPPAs; growth viable post-2030 if renewables ≥80%.
  • Regional diversification + co-location eases the Dublin grid.

Caveats & limits

  • Explicitly not a cost-benefit analysis — costs assessed more narrowly than benefits.
  • Construction jobs projected to fall toward zero without new grid contracts.
  • Further expansion still raises carbon until firm low-carbon backup scales.
  • Government-commissioned; flags knowledge gaps on environmental/social data.

UNU-INWEH — the environmental-footprint case

Claims

  • AI's true footprint must be measured in carbon, water and land — not carbon alone; a "low-carbon" grid can still be high-water or high-land.
  • Global DC electricity (448 TWh in 2025) and all three footprints roughly double by 2030 on current trends.
  • Inference, not training, dominates: ~80–90% of AI energy use comes from everyday queries at scale.
  • Efficiency gains alone won't reduce totals — the rebound effect (Jevons Paradox) can erase them.
  • Burdens fall locally (water stress, land, grid) while benefits flow to distant host firms — an equity/justice problem; cites Ireland as a cautionary case.

Scope & limits

  • Global and AI-centred, not an analysis of the Irish data-centre sector specifically.
  • Footprints derived from global average intensity factors and literature benchmarks — estimates, not measured site data.
  • Training-energy figures for frontier models (e.g. GPT-4/5) are independent estimates, not disclosed.
  • Explicitly not a cost-benefit analysis; does not quantify AI's economic value or offsetting benefits.
Head-to-head: DII and KPMG treat data centres as a value engine and the energy issue as a solvable delivery/alignment problem. Friends of the Earth Ireland treats the same demand as the mechanism that socialises gas-price risk onto Irish households. UNU-INWEH zooms out to the global, AI-driven picture and adds the dimensions the others omit — water and land — arguing even renewable-powered growth carries hidden environmental and equity costs. The four reports largely don't refute each other's numbers; they measure different ledgers — value enabled, monetary cost, and environmental cost — which is why each can be internally consistent yet point toward different policy conclusions. The common thread all four share is the same scale fact: data centres are now a structural, fast-growing share of electricity demand.

What each report concludes

The bottom line of each document, in its own logic — and the specific actions each recommends.
Digital Infrastructure Ireland

Conclusion: act now or lose the investment

DII report cover
€18bninvested in Ireland's DC ecosystem to date
105facilities across 35 operators
€107.5bnICT GVA — 21% of the economy
57%of corporation tax from just 10 firms

Data centres are critical national infrastructure — "the motorways of the 2020s." The danger is not their growth but Ireland's inability to power it: delays risk pushing mobile IP assets, FDI and future corporation tax abroad. Demand exists; the question is whether Ireland can deliver the conditions to capture it.

Recommends
A clearly defined National Digital Strategy aligning energy, planning and industrial policy; faster offshore wind and grid build-out; treating digital infrastructure as critical infrastructure.
Friends of the Earth Ireland

Conclusion: growth socialises cost onto households

Friends of the Earth / Fearon report cover
€360avg household cost, 2015–23 (€715m total)
€644avg household cost to 2034, high-demand case
8.5%of the average bill in 2022's price spike
€1.6bnnational cost to 2034 with a price shock

Inflexible, always-on data-centre demand pushes the gas-dependent grid up the merit order, raising wholesale prices borne by households — an estimated €360 on average over 2015–23, and €295–€644 more per household over 2025–34 depending on demand (up to €1.6bn nationally with a price shock). The effect is regressive.

Recommends
Cap data-centre energy demand and pursue more ambitious renewables — modelled to save the average household ~€435 cumulatively to 2034, even under a price shock.
KPMG (for DETE)

Conclusion: high value, manage the growth

KPMG
The value of data centres to Ireland
€104bnenabled GVA — 19.4% of the economy
876,000jobs enabled across 6 dependent sectors
€2.2bndirect DC GVA · 19,500 jobs (2024)
€2.1bnIrish DC construction exports — 40% of HTC

Data centres underpin substantial economic value — ~€104bn enabled GVA, 876,000 jobs, €2.2bn direct GVA and a "sticky" FDI footprint. Stagnation risks eroding competitiveness and constraining public services; continued growth offers further gains. Explicitly not a cost-benefit analysis.

Recommends
Regional diversification and co-location with renewables (CPPAs); managed — not halted — Dublin growth; post-2030 expansion viable if renewables reach ≥80%.
UNU-INWEH (UN)

Conclusion: govern the full footprint

UNU-INWEH report cover
399 MtCO₂e from global DCs by 2030 (×2 vs 2025)
9.3tn Lwater footprint by 2030 (×2)
14,500 km²land footprint by 2030 (×2)
945 TWhglobal DC electricity by 2030 — ~3% of world

AI-driven data-centre growth roughly doubles global carbon, water and land footprints by 2030. "Low-carbon" is not automatically low-water or low-land, and efficiency gains can be erased by rebound (Jevons Paradox). Burdens fall locally while benefits flow elsewhere — an equity problem, with Ireland cited as a cautionary case.

Recommends
Six principles — transparency, efficiency-by-design, equity/justice, lifecycle responsibility, global cooperation, sustainable use; standardised footprint disclosure and resource budgets.
Read together, the conclusions are less a contradiction than a layering: DII and KPMG conclude the economic case favours managed growth; Friends of the Earth Ireland concludes the household-cost case favours caps; UNU-INWEH concludes the environmental case demands full-footprint governance regardless of where growth happens. All four accept the underlying scale trend — they differ on which ledger should drive the decision.