In Weida recently released a warm-water cooling system and stated that it could significantly reduce water use on site in data centres. According to external sources, this statement is based primarily on the interior of the room; the total water pressure of the AI data centre has not been really addressed by including power generation and chip manufacturing.
Water in the room is close to zero.
According to In Weida, the system uses closed-cycle cooling fluids, which can be repeated over the life cycle of the facility after first injection and no longer sustain new water consumption. According to the company, on-site cooling water can be reduced by 100 per cent in areas with more suitable climatic conditions.
It works by bringing cooling fluids about 45 degrees centigrade into the cabinet, rising to about 55 degrees centigrade through the server and bringing heat out of the hardware. Owing to high temperatures, many areas can be reduced by direct use of outdoor air and passive heaters for cooling and evaporation cooling, and, in part, by fan and chiller use.
It's not just the machine room.
According to external sources, the problem is the statistical calibration. Young Weida now emphasizes water performance within the data centre wall, where electricity generation and chip production are also important sources of water consumption for AI infrastructure.
The report cites studies that, if these links are included, the total water footprint of the data centre may be twice to twice as high as that of the inside of the single-watch room. This means that the new programme in In Weida may have addressed only about one quarter to one third of the total water use in AI data centres.
Power sources determine external water consumption
It is mentioned that as long as AI data centres remain heavily dependent on fossil fuels for power generation, external water consumption is unlikely to decrease significantly. Data from the United States Geological Survey indicate that US fossil fuel power plants consume approximately 2.7 billion gallons per day, mainly for evaporation cooling.
A recent study shows that natural gas power consumes approximately 1.17 litres per kilowatt hour and coal-fired electricity is about 2.2 litres. According to International Energy Agency data, about half of the data centre electricity remains from fossil fuels.
In contrast, water use for wind and solar power is significantly lower. It is reported that wind power consumes approximately 0.01 litres of water per kilowatt-hour and solar energy about 0.03 litres, a figure that already includes equipment manufacturing and solar panels cleaning.
According to external sources, reducing total water consumption in AI data centres depends not only on the cooling system itself, but also on whether the power structure is shifting to low water-consuming energy. TechCrunch states that evidence has been obtained from British Weida regarding the statistical caliber.