People often object to AI's energy use simply because it represents a *new* source of emissions. This psychological bias distracts from the fact that these new emissions are minuscule compared to massive, existing sources like personal transportation.

People focus their environmental efforts on highly visible but low-impact items like plastic bags and recycling. The climate and environmental impact of the food products they purchase—particularly meat—is orders of magnitude greater. This reveals a massive misallocation of public concern and effort.

Although 90% of an AI server's financial cost is the upfront hardware purchase, the vast majority (~95%) of its lifetime carbon footprint comes from the electricity used to run it, not from its manufacturing.

The most effective climate action for transport is not replacing every gas car with an EV, which trades one problem set for another. The superior solution is redesigning cities for walkability, cycling, and public transit to reduce the total number of vehicles needed in the first place.

Citing a Harvard Business School study of 1,800 companies, Sir Ronald Cohen reveals the staggering scale of negative externalities. A third of these firms (600) cause environmental damage equivalent to a quarter or more of their profits, while 250 create more damage than they make in profit, highlighting the financial materiality of impact.

To address its largest emissions source (Scope 3), Mars goes beyond farmer engagement by re-engineering its products. Scientists use a dedicated IT system to reformulate pet food, optimizing for both animal nutrition and environmental impact by prioritizing lower-impact proteins like poultry over beef, tackling the problem at the R&D level.

The model of pressuring tech companies to go green doesn't apply to major industrial emitters like oil and steel. For them, the cost of eliminating emissions can be several times their annual profit, a cost no shareholder base would voluntarily accept.

A single 20-mile car trip emits as much CO2 as roughly 10,000 chatbot queries. This means that if AI helps you avoid just one such trip, you have more than offset a year's worth of heavy personal AI usage.

When sustainable investors starve "brown" (high-emission) companies of capital, those firms become capital-constrained, which can lead them to increase emissions. Meanwhile, investing more in already-green firms has little impact on their already-low emissions. The net result of this common ESG strategy could be an overall increase in pollution.