Oasis is a concept-stage thermal system for AI data centers. It captures the heat GPUs pour into cooling loops, buffers the spikes, reuses what it can — and cuts the water that evaporative cooling drinks.
Proceeds from the OASIS token fund this research, published in the open.
Connects to existing cooling loops or liquid-cooled racks and captures the heat GPU workloads leave behind.
System / 01 ↗Stores and smooths thermal load, so sudden training spikes stop forcing the cooling plant into its worst hours.
Simulator / 02 ↗Routes usable heat toward recovery processes instead of venting it — waste heat treated as a resource.
Research / 03 ↗Rejects what remains with dry or low-water methods wherever the climate allows, trimming evaporation only when it must.
Simulator / 04 ↗U.S. data centers drew 17.4 billion gallons of water directly in 2023, and roughly two-thirds of those built since 2022 sit where water is already scarce. Oasis explores a retrofit that moves as much heat rejection as the climate allows onto dry, closed-loop paths — modeled, not yet measured.
Every megawatt a GPU draws becomes a megawatt of heat. One site in Denmark already pushes about 45 MW of it into a city heating network, warming 12,000 homes — so recovery is engineering, not speculation. System 001 captures heat at the loop, smooths it through a thermal buffer, and routes what it can toward recovery before rejecting the rest.
An experimental thermal system for AI data centers. Proceeds from the OASIS token fund the research, and the models, results and progress are published in the open.
Concept stage. Performance figures are modeled,
not measured in an operating facility.