Maxwell's Demon

Maxwell’s demon is a thought experiment that the Scottish physicist James Clerk Maxwell formulated in 1867. Imagine a container of gas divided into two halves by a wall. In the wall sits a small hatch, and next to it sits a tiny creature, the so-called demon. The gas particles fly around at different speeds, and the demon observes each and every one of them. He opens the hatch only when a fast particle wants to move right or a slow one wants to move left. After some time, all the fast particles collect on the right and all the slow ones on the left – one side becomes hot, the other cold, without any heating at all. This is exactly what should be impossible according to the laws of thermodynamics.

The Attack on the Second Law

The second law of thermodynamics is one of the most stable rules in physics. Put simply, it states: heat always flows spontaneously from hot to cold, never the other way around. A cup of coffee cools down; it does not heat itself back up on its own. Disorder increases in a closed system, never decreases. The demon seems to break exactly that, since he creates an ordered hot and a cold half out of a uniformly lukewarm gas.

This would not merely be a neat paradox, but economically enormous. Work can be extracted from a temperature difference, for example with a turbine. A functioning demon would thus be a machine that continuously delivers energy without consuming fuel. Physicists call something like this a perpetual motion machine of the second kind. For over a hundred years it was unclear exactly why this doesn’t work.

The answer changed our understanding of what information actually is. It is not a disembodied idea, but something that requires space, costs energy, and is subject to physical laws. This thought directly connects the thermodynamics of the 19th century with the computer science of the 21st.

Why the Demon Pays After All

The first suspicion was obvious: the demon must somehow see the particles, and measuring costs energy. But this explanation falls short, since in theory measurement can be made arbitrarily cheap. The decisive point lies elsewhere. The demon must remember which particle is coming in order to decide correctly. He is thus accumulating information, and his memory is finite.

At some point the storage is full and must be erased. This is exactly where Landauer’s principle comes into play, named after the physicist Rolf Landauer. It states: erasing a single bit, i.e. the smallest unit of information, inevitably releases a tiny amount of heat into the environment. This heat is exactly large enough to consume the order that the demon has created. In the end, the second law remains untouched.

One can picture it like a bookkeeping account. The demon makes a profit on the gas side, but in doing so accumulates debt in his memory. When the memory is cleaned out, the debt comes due, and the balance ends up at zero or worse. The Landauer limit has since been experimentally confirmed several times, among other things with individual particles in laser traps.

From Thought Experiment to Electricity Bill

The Landauer limit is today a real physical lower bound for computers. At room temperature it is about 3 times 10 to the power of minus 21 joules per erased bit. That is absurdly little, but it is not zero. Current chips still consume a thousand to a million times that amount per computing step. The demon thus marks the limit that the chip industry has been approaching for decades.

In reports about AI, this connection usually appears indirectly. When it is reported that data centers need as much electricity for training large models as small towns, the same physics lies behind it: computing generates heat, and heat must be cooled away. Research on reversible computing, in which bits are ideally never erased, directly builds on the demon.

A common misconception is to regard the demon as a failed invention. He was never meant as a blueprint, but as a thinking tool. His value lies in the fact that his refutation took almost a hundred years and in the process forged a new connection: between heat, order, and information. Without this connection, there would be neither modern information theory nor a clear understanding of why computing fundamentally costs energy.