The cost of cheap energy

Every device in a school hall runs on electricity that came from somewhere, and where it came from has changed faster in the last decade than in the previous century. Solar panels and wind turbines, once the most expensive way to generate power, are now among the cheapest in most of the world. The reason is not a single invention but a long, boring process: each doubling of production brought a predictable fall in unit cost, repeated over thirty years.

That progress creates a new problem rather than ending the old one. Sunlight and wind are variable. A grid must match supply and demand second by second, so a system with a large share of renewable generation needs something to cover the evening, the still week, the cloudy month. Batteries handle hours well and seasons badly. Hydropower can store energy where geography allows. Gas plants, built to run constantly, are increasingly used as expensive insurance.

The debate that follows is often framed as a choice between technologies, which is the least useful way to think about it. The binding constraints are rarely scientific. They are the speed at which transmission lines can be approved, the availability of trained electricians, the price of borrowing money for projects that pay back over twenty years, and the willingness of a community to accept a substation near its houses.

Schools sit inside this system in a modest but instructive way. A classroom's consumption is small; its influence is not. Students who understand why the lights dim when a large motor starts, who can read a meter and calculate the cost of leaving thirty computers on overnight, make better decisions as adults than students who have only been told that energy should be saved.

There is also a lesson about design. A platform that runs its artificial intelligence on each student's own device uses the battery of a tablet rather than the power of a distant server farm. That choice was made for privacy and cost, not for the climate, and the saving is far too small to matter on its own. It is worth stating anyway, because it illustrates a general rule: the environmental cost of software is decided by architecture, long before anyone is asked to switch off a light.