Cloud City

AI computing built at its power source: data centers grown along a geothermal rift in Iceland.

We chose the site with a multi-criteria matrix weighing geothermal yield, visibility, slope and protected land; Boone led the GIS mapping that found an active rift corridor. Growth was then modeled in Fieldwork, a simulation we built on a 250 m grid along 78 km of that corridor, where facilities spread according to server heat, free-cooling limits and how much water the wellfields can take back.

My part was the geothermal and data center side. I traced a single AI image request from a phone to a GPU and back, and what it costs in energy and material; drew how the server rack changes from 2025 to 2050; and stacked the building's cycle floor by floor, from cycling and generating at the base to processing and computing above, over wells drilled kilometres down.

It is assembled from a catalog of modular steel parts and connections taken from offshore platforms, so it can stand over the wells with little disturbance at the surface.