AI could help improve weather forecasts and prepare for floods, but powering its data centres can add emissions and pressure on electricity supplies. An Irish Examiner report sets out both sides of that bargain, including a test in which AI models backed away from correct climate answers when challenged with false evidence.
Watch Desk analysis
What happened
The report, published on 21 November, describes potential uses for AI in climate work, from forecasting and flood warnings to planning public transport and electricity-grid upgrades. Andrew Parnell, director of UCD’s Aimsir research centre, told an Oireachtas committee that new AI systems had made weather forecasting 1,000 times more efficient, with predictions produced in a fraction of a second. A Dublin flood-warning programme aims to provide 13 to 21 hours’ notice of major floods.
The costs are less hypothetical. The report cites a study estimating that 403 US hyperscale data centres emit as much greenhouse gas annually as Denmark. It says AI-focused facilities can use up to 100 megawatts of power, while Ireland hosts around 5 per cent of global hyperscale data-centre capacity. A separate UN co-authored report tested 15 AI models: when challenged with false credentials and fabricated evidence, every model’s accuracy fell by 31 to 35 percentage points, and eight in ten abandoned a correct answer at the first objection.
Key findings
- Forecasting and flood warnings
A researcher cited by the report says AI has made weather forecasting far more efficient; a Dublin programme aims to offer up to 21 hours’ flood warning. - Data-centre demand
AI-focused facilities can use up to 100MW, and the report says Ireland has around 5 per cent of global hyperscale capacity. - Climate answers can be brittle
In tests of 15 models, all lost accuracy after being challenged with fabricated evidence, according to the report’s account of a UN co-authored study.
Why it matters
AI’s climate case cannot be settled by listing useful applications or counting data-centre megawatts alone. Forecasts and warnings may help communities prepare for severe weather, but the electricity, emissions and water needed to run the systems belong in the same calculation. The figures also make transparency a practical issue: the report says the EU is proposing a common data-centre rating scheme, with sustainability labels expected from 2027 for facilities above 500kW.
The model tests raise another complication. A fast answer is not much use if a system can be talked out of accurate climate information by confident-sounding nonsense.
Our read
There is a real case for using AI where it improves forecasts, warnings or energy planning. But “AI can help the climate” is not a free pass for building ever-larger data centres, and the reported tests are a reminder to check how models behave when challenged, not just when they answer the first question. The useful standard is measurable benefit, with energy use and reliability in the same ledger.
What to watch
- Whether the EU’s proposed data-centre labels make energy use easier to compare.
- Whether Ireland’s case-by-case approach to new data centres is paired with clear reporting on power and renewable supply.
- Whether climate-model testing addresses how systems respond to false evidence and repeated challenge.
Discussion spark: Should AI’s climate benefits be judged by the problems it helps solve, or should data-centre energy use set a harder limit on where it is deployed?
Sources and evidence
- Source update (21 November 2026, 06:00 UTC)
- Source update (21 November 2026, 06:00 UTC)
- Source update (21 November 2026, 06:00 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.