See what's public. Know what isn't.
Candor tracks the data centers behind AI and the grids that power them. It shows what the public record leaves out. And it investigates how better data could help businesses and people choose where, when and how AI runs.
AI runs on electricity from specific facilities,
on specific grids, at specific hours.
Much of that isn't disclosed.
Candor's review in October 2026 found that, among the providers it checked, one has published a measured energy figure for a typical prompt. For the others, the public has statements without methods, and independent estimates that differ from one another by several times. For many facilities there is no settled public record of who uses them.
Without that data, a business deciding where and when to run AI workloads, a company choosing between AI services, and a person choosing between tools have little evidence to compare their options on energy.
The challenge isn't only measuring AI's footprint. It's making the data available that lets people decide what changes it.
What data exists about AI's electricity use, what doesn't, and why. Mapping facilities, grids and disclosures, and stating the gaps plainly.
What could actually change the picture. Whether timing, location, model choice and efficiency measurably change energy use and emissions, and whether those same choices cost less to run. The cost question has not been researched yet.
How better data could help businesses and people choose where, when and how AI runs, and inform organizational accountability and policy, with evidence anyone can check.
Candor began with a question: can the infrastructure supporting artificial intelligence be connected to the electricity systems that supply it?
The Candor Infrastructure Tracker shows where AI data centers are, what grid each one draws from, and how clean that grid has been, hour by hour. It also shows where the public record doesn't say. In the data so far, the cleanest hours come at different times of day on different grids, and by different amounts. It's one research instrument Candor has built, not the whole of what it investigates.
Every figure's source, formula and assumption is published in full, including the parts that are estimates and not measurements.
Data alone doesn't change anything. Decisions do. Knowing how much electricity AI uses, and when and where that electricity is cleanest, could help businesses and people choose differently.
Candor is also investigating whether those choices cost less to run. That work is still ahead, and this page says so. Candor's ambition is to contribute evidence that makes better decisions possible.
Clear data.
Better decisions.