Artificial intelligence is transforming how we work, build, and innovate. But understanding its environmental consequences — and determining how to reduce them — remains an evolving scientific and technical challenge.
Candor Intelligence investigates the intersection of AI infrastructure, electricity consumption, environmental transparency, and evidence-based decision-making.
We cannot improve what
we do not adequately understand.
AI's environmental impact extends beyond the electricity consumed by a single prompt. It involves computational efficiency, infrastructure utilization, electricity generation, geographic location, and the decisions that determine how AI systems actually operate.
Yet critical information remains difficult to access, measure, or independently verify.
The challenge isn't simply measuring AI's footprint. It's understanding which decisions can meaningfully change it.
What is AI's environmental impact? Investigating infrastructure, energy consumption, electricity grids, and the limitations of current environmental reporting.
What can actually reduce it? Examining computational efficiency, model selection, inference optimization, and electricity-related interventions.
How can evidence shape better decisions? Exploring how environmental research can inform technical practices, organizational accountability, and policy development.
Our investigation began with a fundamental question: can we connect the infrastructure supporting artificial intelligence to the electricity systems supplying it?
The Candor Infrastructure Tracker brings together publicly available facility information and electricity-grid data to explore what can be measured — and where important uncertainties remain. It's one research instrument we've built along the way, not the whole of what we're investigating.
Knowing how much electricity an AI system consumes is valuable. Knowing how to reduce that consumption — and whether doing so actually reduces emissions — is a different challenge.
Candor's research is evolving beyond environmental accounting toward investigating the relationship between measurement, intervention, and meaningful environmental outcomes. Our ambition is to contribute evidence that makes better decisions possible.
Better intelligence.
Less impact.