Your AI.
Without water
to cool it.

The proposed configuration places LLooM in your infrastructure, on air-cooled hardware assessed for the workload, such as a Mac mini or NVIDIA DGX Spark. Directly cooling the device during use requires no water.

0L

of water needed to cool the device directly

This covers only direct cooling of the device during use. Electricity, hardware manufacturing and room air conditioning are outside this calculation. It does not represent the total water footprint.

Cooled by air. Not by water.

Mac mini and DGX Spark release heat into the air: directly cooling these devices does not require a water circuit. The model runs in your infrastructure.

Cloud computing can also use water.

In data centres with evaporative cooling, some water is consumed to remove heat. AI requests contribute to the facility's workload; water use depends on the cooling system and the work performed.

Cloud services use different cooling systems. This comparison does not measure the total water footprint or savings relative to the cloud.

What does a day of use consume?

An estimate based on your assumptions, not measured LLooM consumption.

0.88kWh / day

0.48 kWh under load + 0.40 kWh in the other 20 hours.

How the estimate is calculated

E = [W under load × hours under load + W during the remaining hours × (24 − hours under load)] / 1,000. This excludes room air conditioning, other devices, hardware manufacturing and model training. It does not estimate CO₂ or water use, or demonstrate an advantage over the cloud. Import activity must be measured; it cannot simply be assumed to fall within these inputs.

DETAILS