New suggested way to estimate AI emissions

A software company called Watershed has proposed a way for businesses to estimate the greenhouse gas emissions tied to their use of artificial intelligence. Right now, companies that build AI tools, like Anthropic and OpenAI, mostly don't share detailed data on the energy and emissions behind each AI query. Big cloud providers such as Amazon, Google, and Microsoft have released some figures, but the details are still thin. Watershed's proposal suggests measuring emissions per million "tokens," the small chunks of text that make up an AI prompt, along with the electricity used to generate them.

This is mainly a story about how businesses track their own carbon footprints, not something that changes anything about your house or your energy bills. But it connects to a bigger trend worth knowing about: AI systems run on data centers that use large amounts of electricity, and that demand is growing fast enough to affect regional power grids. As more companies try to measure and disclose this impact, it could eventually shape how utilities plan for demand and price electricity, which does trickle down to household rates over time.

For now, there's nothing here that calls for you to do anything to your home. It's a look at how the tech and corporate sustainability world is trying to get a handle on AI's hidden energy costs, a preview of decisions that may, eventually, ripple out to the grid we all depend on.