“Irreversibility and Heat Generation in the Computing Process,” by Rolf Landauer of IBM, appeared in the IBM Journal of Research and Development in 1961, volume 5, pages 183 to 191. It established a deep and surprising link between abstract logic and physical thermodynamics, a result now known as Landauer’s principle.
Landauer’s insight was that erasing information is not free. When a logical operation throws away information - for example, resetting a memory bit to zero regardless of its previous value, so that two possible states collapse into one - that loss is logically irreversible, and logical irreversibility forces physical irreversibility. By the second law of thermodynamics, it must be accompanied by an increase in entropy somewhere, which means a minimum amount of energy must be dissipated as heat. He put that floor at roughly kT ln 2 per erased bit, where k is Boltzmann’s constant and T is temperature - about 3 times 10 to the minus 21 joules at room temperature.
The number is almost unimaginably small; today’s chips dissipate millions of times more per operation than the Landauer limit demands, so the bound is nowhere near the practical constraint on computing. But it is a hard floor set by physics, not engineering, and it cannot be beaten by any conventional irreversible logic. It also pointed toward reversible computing - the idea that computations which never erase information could in principle dissipate no heat at all - a thread later developed by Charles Bennett and others.
Why business readers should care: the energy cost of computation is now central to AI economics and even to national power-grid planning. Landauer’s principle is the bedrock statement of why computing costs energy at all, and the ultimate reference point for how far more efficient AI hardware could theoretically go.