The earth is being drafted into the AI arms race. Google's 400 MW deal with Fervo Energy, potentially expandable to a full gigawatt, is not a climate story. It is an infrastructure story wearing a climate story's clothes. A gigawatt is enough to power a very large AI data center in Utah. The UN just released what it called one of the bleakest climate reports in its history, acknowledging that Earth has lost its chance to prevent some warming. Meanwhile Chevron is expanding oil operations in Venezuela. The needle is not moving in the direction the press releases suggest.
Enhanced Geothermal and the AI Energy Problem
Enhanced geothermal is not your grandparent's geothermal. Traditional geothermal power only works where the earth's heat is conveniently close to the surface, basically Iceland and a few zip codes in California. Enhanced geothermal drills deep, fractures rock, and circulates water through the cracks to harvest heat anywhere. Fervo is the company that proved this works at commercial scale, and Google is now its anchor customer. The logic is elegant: geothermal is always-on, unlike solar or wind, which is exactly what a data center needs. AI training runs do not pause for clouds. The problem is that building these plants at the scale AI demands will take years, and the models are training now, on grids still threaded with coal.
The Climate Cognitive Dissonance at the Heart of Big Tech
Google is simultaneously partnering with MrBeast to market Gemini at scale, defending its ad tech empire in court, and buying clean power. The posture is: we are building the thing that requires the energy, and we are also solving the energy problem. A 2023 paper in Nature Communications by Luccioni et al. found that generative AI inference, the part that runs every time you ask a model a question, consumes orders of magnitude more energy than traditional search. The Fervo deal is real and good. It is also a rounding error against what comes next if Astra, OpenAI's forthcoming flagship model, and its peers achieve the adoption rates their makers are projecting. The question is not whether geothermal is good. It is whether any supply can keep pace with demand that has no ceiling anyone is willing to name.