Energy, AI, and Bitcoin: Three Threads
Building energy analytics, AI tooling, and Bitcoin's energy narrative. Why I think they're converging.
Three things I spend most of my time thinking about are starting to overlap in ways that feel important.
Buildings are data problems
Commercial buildings waste 30% of the energy they consume. Not because the technology to fix it doesn’t exist, but because the data is fragmented, unlabeled, and locked in proprietary systems. My day job is building tools that make this data legible, so engineers can actually find and fix the waste.
AI is making the tools cheaper
The models I used in my PhD research three years ago (BERT, topic modeling, custom NLP pipelines) required weeks of setup and domain expertise. Today, I can get 80% of the same results with an API call. That’s not hype; I’ve measured it. The implication is that smaller teams can tackle problems that used to require research labs.
Bitcoin forces honest energy accounting
Bitcoin mining is one of the few industries that pays the actual marginal cost of electricity, in real time, everywhere. That’s the start of a real-time energy market: a global price signal that responds to curtailment, transmission constraints, and oversupply as they happen. Grids today still run on day-ahead forecasts and bilateral contracts. Mining is the closest thing we have to demand that actually clears at the margin.
The next layer is more interesting. The same hardware and the same cheap stranded power that mines bitcoin can also run AI inference. You’re effectively trading energy tokens for AI tokens, with bitcoin’s price discovery sitting underneath both. The cost of generating an LLM token has less to do with model architecture than people think. Mostly, it’s about where and when the electrons came from.
The debate about whether Bitcoin “wastes” energy is the wrong frame. The better question is whether the energy accounting it forces, and the markets that grow on top of it, will change how we plan and operate grids.