Imagine this: I’m enjoying myself at a gathering, stationed near the snack area (naturally). A connection of a friend approaches, and we engage in dialogue. The conversation swiftly shifts to work, and upon discovering my role as a climate technology journalist, my new acquaintance asks something akin to: “Is using AI a bad idea? I’ve heard it harms the environment.”
This occurs quite frequently nowadays. Typically, I advise people not to stress—let a chatbot organize your trip, propose meal ideas, or compose a poem if that’s what you wish.
This response may catch some off guard, but I assure you, I’m well-informed, and I’ve encountered all the alarming forecasts regarding AI’s electricity consumption. Data centers could use as much as 945 terawatt-hours annually by 2030. (That’s roughly equivalent to Japan’s consumption.)
However, I firmly believe we shouldn’t place all the responsibility on individuals, partly because AI discussions remind me so much of another question: “What can I do to lessen my carbon footprint?”
That question annoys me due to its background: BP helped popularize the notion of a carbon footprint in a marketing push during the early 2000s. This perspective effectively transfers the responsibility of environmental concerns from fossil fuel companies to individuals.
The truth is, no single person can tackle climate change independently: Our whole society is structured around consuming fossil fuels. To confront climate change, we require political action and public backing for advancing and scaling climate technology. Companies must innovate and take pivotal steps to cut down greenhouse-gas emissions. Overemphasizing individual actions diverts attention from the genuine solutions available.
I notice a similar trend today concerning AI. People are inquiring of climate journalists at social events if they should feel guilty for frequently using chatbots when we ought to fixate on broader issues.
Major tech firms are contributing to this narrative by supplying energy consumption estimates for their products at the user level. Several recent reports estimated the energy consumed to interact with a chatbot at about 0.3 watt-hours, which is equivalent to running a microwave for roughly a second. That’s so minimal it’s practically negligible.
However, ceasing with the energy usage of an individual inquiry obscures the larger picture, which is that this sector is expanding rapidly, constructing energy-intensive infrastructures at an almost unimaginable scale to fulfill society’s overall demand for AI. Meta is currently establishing a data center in Louisiana boasting five gigawatts of computational capacity— comparable to Maine’s total demand at peak summer times. (For additional information, refer to our Power Hungry series online.)
Increasingly, there’s no escaping AI, and it’s not simply a matter of opting to use or avoid the technology. Your preferred search engine likely provides an AI summary at the beginning of your search results. The suggested replies from your email provider? Probably AI-generated. The same applies when communicating with customer service during online shopping.
Similar to climate change, we must view this as a system rather than a sequence of personal decisions.
Large technology firms employing AI in their offerings ought to disclose their total energy and water consumption and provide detailed insights into their calculation methods. Estimating the impact per query is an initial step, but we also deserve to understand how these effects accumulate for billions of users, and how that evolves over time as companies (hopefully) enhance the efficiency of their products. Legislators should be enforcing these disclosures, and we should advocate for them too.
This is not to suggest there’s absolutely no individual action you can take. Just as you could make a meaningful dent in your greenhouse gas emissions by taking fewer flights and consuming less meat, there are certain reasonable actions you can undertake to minimize your AI footprint. Creating videos tends to be particularly energy-heavy, as does utilizing reasoning models for engaging in lengthy prompts and producing detailed answers. Asking a chatbot to assist in planning your day, recommend enjoyable family activities, or summarize an exceptionally lengthy email has a relatively minor effect.
Ultimately, as long as you aren’t persistently generating AI noise, you shouldn’t be overly concerned about your individual AI impact. However, we should all remain vigilant about what this sector will signify for our power grid, our society, and our planet.
