WattWhere

Percentage of energy from renewable sources, pixel breakdown shown on the right.
Connectivity score, a combination of internet, broadband, roads and grid access.
Cooling score, a measure of the amount of available water in the area which is used for cooling the data centre.
Various bodies of water in County Donegal.

Inspiration

Global data centre capacity is projected to nearly double by 2030, from 103 to 200 GW. Picking where to build data centres is a difficult, time-consuming process, and most importantly, when handled incorrectly, can have drastic consequences for the environment and local communities.

The bad press around the recent ramp-up in the construction of AI data centres includes stories of nearby residents not being able to drink tap water anymore, and reports that the 'AI boom' has caused as much carbon dioxide to be released into the atmosphere in 2025 as emitted by the whole of New York City.

This doesn't have to be the case.

What it does

WattWhere aims to solve this issue for data centre construction in Ireland, by providing an interactive map which allows users to click through ~14,000 grid squares across the entire country and group key metrics for data centre viability in each region by 'sorts'.

Each tile can be explored through six sorts:

Energy: Renewable generation potential (wind speed at 100m hub height, solar irradiance) combined with proximity to transmission lines and substations, rewarding regions with clean, accessible power.
Environmental Constraints: Overlap with protected designations (SAC, SPA, NHA) and flood risk zones. Tiles intersecting Special Areas of Conservation, Special Protection Areas, or current flood extents receive an automatic hard exclusion.
Cooling & Climate: Ambient temperature (lower is better for free-air cooling), proximity to rivers and lakes, and annual rainfall, reflecting the enormous water demands of modern data centre cooling systems.
Connectivity: Broadband coverage quality, distance to Ireland's internet exchange points (INEX Dublin and Cork), and road/motorway access for construction logistics and ongoing operations.
Planning & Zoning: Local development plan classifications, proximity to existing industrial and enterprise zones, IDA Ireland-designated sites, and prior planning applications for data centre use.
Overall: A configurable weighted composite of all five dimensions (default: Energy 25%, Connectivity 25%, Environment 20%, Cooling 15%, Planning 15%), giving a single at-a-glance suitability score.

Each sort also exposes sub-metrics. For example, within Energy, you can switch the map colouring from the composite score to raw wind speed, solar irradiance, or grid proximity alone, which allows users to drill into the specific factor driving a region's performance.

Furthermore, we calculate a score for each region between 0 and 100, derived from real Irish government open data: Global Wind Atlas rasters, Met Éireann climate grids, NPWS protected area shapefiles, OPW flood maps, ComReg broadband coverage, CSO census small-area statistics, and MyPlan zoning data. All scoring is deterministic and stored in a PostGIS database, with raw values normalised at tile-serve time so the map always reflects the full dataset range.

For each selected tile, WattWhere also generates an AI executive summary — a concise, plain-language verdict written for a non-technical audience. Each of the five sort categories gets its own tailored summary highlighting the factors that matter most (e.g. grid viability and renewable potential for Energy, or Natura 2000 hard exclusions for Environment). The Overall summary pulls together data from every sub-category to deliver a holistic site recommendation, flagging the strongest and weakest dimensions and any deal-breaking constraints in two to three sentences.

How we built it

WattWhere is a full-stack geospatial application built around a PostGIS database containing one row per tile (roughly 14,000 in total).

A Python data pipeline ingests over a dozen raw datasets from Irish and international open-data sources, spanning multiple formats: GeoTIFFs from the Global Wind Atlas and NASA POWER (wind speed, solar irradiance, temperature, rainfall), Shapefiles from NPWS (SAC, SPA, NHA protected area boundaries) and OPW (national flood extent maps), ArcGIS REST feature services from GSI (landslide susceptibility, bedrock aquifer maps) and ComReg (broadband coverage polygons), CSV extracts from SEAI (connected wind farms) and the Property Price Register, and OpenStreetMap Overpass queries for power infrastructure, road networks, and waterways. The pipeline first generates a ~14,000-tile grid over Ireland in EPSG:2157 (Irish Transverse Mercator), then downloads and processes all five sort categories in parallel. All scores are upserted idempotently into PostgreSQL so the pipeline is safe to re-run, and raw values are stored un-normalised. Normalisation happens at tile-serve time so the map always reflects the full dataset range.

A Martin tile server dynamically generates Mapbox Vector Tiles from PostGIS functions, accepting sort and metric query parameters to drive map colouring without any pre-rendering. This makes sort and sub-metric switching near-instant.

A FastAPI backend exposes REST endpoints for tile detail data, pin locations, score breakdowns, and an admin endpoint for adjusting composite weights.

The Vue 3 frontend uses MapLibre GL JS for rendering, Pinia for state management, and sort-specific sidebar components that surface the raw underlying data for any selected tile: nearest substation name, hydrometric station flow rates, planning application references, and more.

An Anthropic Claude API integration (Claude Sonnet 4.6) powers per-tile executive summaries. The FastAPI backend collects all relevant metrics for a selected tile and sort category, then calls the Claude Messages API with domain-specific system prompts tailored to each category (energy, environment, cooling, connectivity, planning, or a unified overall prompt that aggregates data from all five dimensions). The result is a 2–3 sentence plain-text summary aimed at a non-technical executive audience.

Everything is containerised with Docker Compose, with nginx acting as a reverse proxy in front of the API and tile server.

Challenges we ran into

Electrical grid data was the most significant gap in Ireland's open data ecosystem. Unlike the environmental, flood, and climate datasets, which were all available from official Irish agencies with clear licensing, no authoritative open shapefile exists for the transmission or distribution network. We fell back on OpenStreetMap, which required extra validation to confirm coverage quality.

Pipeline efficiency was an early bottleneck. Our initial approach downloaded large GeoTIFFs and shapefiles in pure Python, buffering entire files into memory before writing to disk: slow and fragile on files in the tens-of-megabytes range. Switching to wget for the heavy lifting cut download times dramatically and made the pipeline far more reliable.

Accomplishments that we're proud of

Building a tool that makes the environmental cost of infrastructure decisions visible and unavoidable. Protected habitats and flood zones trigger hard exclusions (they simply can't be selected) while the scoring system actively rewards proximity to renewable generation and cooler, wetter climates that reduce a facility's energy and water footprint.

What we learned

The data to make better decisions already exists; it's just scattered. Ireland publishes detailed flood maps, protected area boundaries, wind and climate grids, and broadband coverage, but no single tool brings them together for infrastructure planning.

In general, siting matters far more than it's currently treated as mattering.

What's next for WattWhere

Ireland was chosen because it was our HackEurope base, because Irish open data is good enough to prove the concept, and because data centre provision here is fairly limited.

But the pipeline is country-agnostic by design. Any country with open environmental, climate, grid, and planning datasets (which are many) can be ingested into the same scoring framework. The EU's INSPIRE directive alone means that equivalent datasets exist for every member state.

The data centre boom is not an Irish problem. It's a global one, and the decisions being made right now about where to build will lock in environmental consequences for decades. WattWhere could be the tool that makes those decisions defensible.