Google funds AI irrigation pilot to save water in Belgian farms
TL;DR:
- Google is funding Agua Segura and Agrow Analytics to deploy AI precision agriculture across more than 1,000 hectares of farmland in the Scheldt Basin, the river network running through Belgium.
- The pilot targets up to 158 million gallons (around 600,000 cubic metres) of annual water replenishment by reducing irrigation demand and optimising fertiliser use, using satellite and thermal imagery integrated with climate, water and soil data.
- Resultsense view: this is a familiar play — hyperscalers fund local water projects in regions where their data centres compete for the same resource. Google operates a major data-centre cluster at Saint-Ghislain in the Scheldt Basin, and the optics of “AI saving water” while AI infrastructure consumes it locally will only get sharper.
The Scheldt Basin runs through Belgium and northern France into the North Sea, and is one of the most water-stressed catchments in northwest Europe. Google announced the precision-agriculture pilot through its Belgian sustainability blog on Tuesday.
How the system works
Agrow Analytics’ platform integrates satellite and thermal imagery with climate, water and soil data to recommend per-field irrigation and fertilisation decisions. The model is calibrated for crop type, soil retention and weather forecast, and replaces calendar-based irrigation schedules with usage tied to actual plant and field state. The 158m gallon figure represents the annual water saving Google projects across the participating 1,000-hectare area.
The data-centre context
Google’s Saint-Ghislain campus, in the Scheldt Basin, is one of the company’s largest European data-centre sites and a target for further AI capacity expansion. Across the EU, hyperscaler water consumption has become a sharper political question: Microsoft committed in 2024 to be “water positive” by 2030, and Amazon and Google have both expanded community-replenishment projects in regions where data-centre evaporative cooling competes for groundwater with farms and towns.
Treating “AI saves water for farmers” as a counterweight to “AI consumes water for cooling” is now a standard PR posture. Whether the maths nets out depends on local catchment, the cooling architecture (closed-loop vs evaporative) and the actual water-saving delivered against projections — none of which is trivial to verify.
UK relevance
The UK’s Cluttons-Knight Frank Data Centre Pulse 2025 placed UK data-centre water consumption at the second-highest concern raised by planning authorities after grid capacity. Slough’s water-utility constraints have already delayed at least one major build. The Belgian model — hyperscaler-funded local-agriculture water pilots — is plausibly a template UK operators will be asked to follow as Wales, the West Midlands and Greater Manchester see new AI capacity proposals.
Looking forward
Two things to watch. First, whether Google publishes verified annual water-replenishment numbers from the Scheldt pilot, with independent measurement — vague projections have a poor track record in this space. Second, whether the model gets replicated in UK water-stressed catchments. The Environment Agency’s 2025 catchment-prioritisation list and Ofwat’s Water Resources Management Plans provide the natural frame. UK SMEs in agri-tech and remote-sensing should treat this as a market signal worth tracking.