The War on Heat Is the Wrong War

For half a century, the physics of computation has been a war on heat. Every transistor flip spends energy it cannot recover; every data center pours that energy into cooling towers and radiator fins. The constraint has been treated as a cost — something to engineer around, not to use. A small, young research community is now arguing the constraint is the resource.

Patrick Coles, a physicist at the New York startup Normal Computing, makes the case directly: the field is about designing computers that exploit thermodynamics as a computational resource, not one that fights it. Read against the AI-era power wall — where heat dissipation is now the binding constraint on denser, faster chips — the inversion is the argument. The substrate's most-cited tax is its most underused feature.

The mechanism generalizes. Photonic computing recruits light, which electronics treats as wire-delay. Neuromorphic computing recruits analog noise, which digital logic treats as error. Thermodynamic computing recruits thermal fluctuations — the random microscopic jiggling the second law says energy decays into — and uses that randomness as a search signal, the way a protein folds by falling downhill into its shape.

The Quanta Magazine piece lands the moment the field crosses from "in principle" to "in simulation": a 2025 Nature Communications paper demonstrated thermodynamic computation simulated in standard silicon-based logic circuits, after a community that gathered for its first dedicated conference at the 2019 Computing Community Consortium. The shipping date is not the story. The story is the design space widening: post-Moore hardware is not "smaller transistors." It is different physics recruited as feature.