Arts organisations take their environmental commitments seriously — many have carbon plans, green riders and board-level targets — so it's entirely reasonable to ask: does using AI blow a hole in all that? The public conversation is unhelpfully split between "each question destroys a rainforest" and "it's nothing, ignore it". Both are wrong. Here are the straight answers, with the honest caveat up front: the numbers below are estimates as of early 2026, they're contested, and they're moving. The shape of the picture, though, is fairly stable.
What one query costs
A typical text question to a chatbot consumes a small amount of electricity — current estimates put it in the region of watt-hours in the low single digits, and falling as the technology gets more efficient. For scale: boiling a kettle for one cup of tea uses many times more. If your entire team used AI heavily all year, the direct footprint would be roughly comparable to a handful of car journeys — real, but nowhere near your building's heating, your touring, or your audience's travel, which dominate any arts organisation's actual footprint by orders of magnitude.
Generating images costs meaningfully more than text; video generation costs a great deal more again. A sensible rule of thumb: typing is cheap, pictures are moderate, video is expensive — spend accordingly and don't generate fifty variations for fun.
So the critics are wrong?
No — and this is where honesty matters. The per-query numbers are small; the system is not. Data centres already consume a low single-digit percentage of the world's electricity, the AI boom is driving a construction surge that has electricity grids and planning authorities genuinely worried, training a frontier model consumes energy on the scale of a small town's year, and cooling those buildings uses water — sometimes in places that can ill afford it. The individual guilt framing ("your query costs X") is largely a distraction; the infrastructure question ("what is this industry building, where, powered by what?") is entirely legitimate. Both things are true at once: your usage is a rounding error, and the industry's trajectory is a serious environmental question being decided right now.
What should an arts organisation actually do?
- Don't perform abstinence. Refusing AI on carbon grounds while running a draughty Victorian building and touring by air is not environmentalism; it's theatre, and not the good kind. Keep proportion: act where your footprint actually lives.
- Put AI in your environmental policy at its true scale. A line or two, alongside your position on everything else AI touches — acknowledged, proportionate, not hand-wrung over.
- Ask your suppliers the infrastructure question. This is where an organisation's voice counts for more than its usage. The big providers publish renewable-energy commitments of varying seriousness; when procurement asks pointed questions, the answers slowly improve. You hold suppliers to account on everything else — add this to the list.
- Waste is the sin, not use. Using AI to shrink real work is defensible; churning out hundreds of generated images to pick one, or bolting AI onto things that didn't need it, is the digital equivalent of leaving the lights on. The habit to teach your team is intentionality.
- Reserve the right to update your view. The honest position on AI and the environment has changed twice in three years and will change again. Committing to keep looking is more credible than any fixed stance.
The straight summary
Your Tuesday-afternoon queries are not an environmental decision worth agonising over. The industry being built to serve them is — and the useful response to that isn't individual guilt but collective pressure: proportionate policy, pointed procurement questions, and no waste. Which, an environmentalist might note, was always the answer to everything.
The image above was AI-generated — as ever, we'll always tell you.