WATERSHED
Q. YOU SPENT MORE THAN A DECADE AT EPRI BEFORE JOINING WATERSHED. WHAT MADE YOU LEAVE?
» I’ ve spent two decades studying how policies, technologies and markets alter the energy transition. That work informs utilities and policymakers. But I have increasingly seen companies making decisions with system-level consequences. That includes where to put data centres, what power to buy, which suppliers to work with. Watershed gave me the opportunity to apply the same modelling rigour to those choices.
Q. YOUR WORK HAS RANGED FROM THE INFLATION REDUCTION ACT TO AI’ S ELECTRICITY DEMAND. WHAT’ S THE THROUGHLINE?
» The throughline has been understanding what it actually takes, in physical and economic terms, to change how the energy system works. The underlying strategy is the same across very different contexts: show which assumptions drive the result and how uncertainty affects the decision.
Q. YOU’ VE CONTRIBUTED TO THE IPCC’ S SIXTH AND SEVENTH ASSESSMENT REPORTS AND TWO US NATIONAL CLIMATE ASSESSMENTS. HOW DOES THAT EXPERIENCE SHAPE HOW YOU APPROACH CORPORATE CLIMATE SCIENCE?
» Nothing makes you more sceptical of certainty than spending a few years on assessment work, synthesising across hundreds of studies. A single number may seem definitive but bury differences in data quality and assumptions. Corporate emissions measurement gives companies numbers they can use and shows them how much confidence to place in them. The AI emissions framework helps companies get to a point of action even without perfect data and transparently shows how estimates can improve with better data.
Q. WHY DID WATERSHED DECIDE THE INDUSTRY NEEDED A NEW FRAMEWORK FOR MEASURING AI EMISSIONS?
» Companies have been adopting AI faster than the emissions data are improving. Most companies don’ t yet have a reliable way to measure the emissions from this AI use, which means they cannot set reduction targets or identify decarbonisation levers as emissions grow. Published estimates for the same AI usage can differ by orders of magnitude depending on the system boundary and what you assume about hardware and the grid mix. Waiting for perfect provider data would leave companies blind during a period of rapid growth. We wanted to give sustainability teams a defensible way to start today, and a path to get more precise as better data from providers become available, which is useful for the decision and clear about its limits.
Q. WHAT’ S THE HEADLINE FINDING PEOPLE SHOULD TAKE AWAY FROM THE WHITE PAPER?
» You don’ t need perfect data to calculate and reduce your emissions.
76 October 2026