We call them server farms and then act surprised that they behave like farms. They sit on land. They drink water. They give off heat. They compete with agriculture for the exact same inputs in the exact same watersheds. The metaphor was never decorative. It was a map we refused to read.

Treat the movement of data the way a civil engineer treats the movement of water, as hydrology. Where does it originate, where does it pool, what is the flow rate, and what does it displace downstream? A data center is not a building. It is a diversion in a system that already had users: crops, towns, rivers.

Compute is a crop

Now flip it. If data centers are farms, then compute is a crop, and you can think about it agronomically. Crops have terroir. They have altitudes and climates where they thrive. Cold air is free cooling. Cheap hydro is a growing season. The question is not whether server farms and real farms can coexist. It is where each one grows best, and whether we plant them on purpose or by accident.

• Altitude and cold climate: free cooling, so the water bill drops. Grow heat-dense compute here.
• Abundant hydro or geothermal: cheap steady power. Grow baseload training here.
• Contested watersheds and farmland: the collision zone, where a data center and a field bid for the same aquifer.

The waste heat from a server farm is not garbage. It is warm water looking for a greenhouse. We keep venting it to the sky because we drew the two farms on separate maps.

Shailin Dhar

This is speculative, but it is not soft. The next decade of infrastructure siting is a hydrology problem wearing a technology costume. The regions that win will be the ones that stop treating compute, water, and food as three separate ledgers and start balancing them as one. Plant the compute where the climate already does half the work, and route the heat to where something can grow in it.