Modern society depends on computers, yet even the simplest operations still consume far more energy than theoretical limits suggest. This project asks a simple but fundamental question: why does computation cost so much energy in practice?

Theoretical physics has shown that there is a minimal energy cost for processing information. This limit is known as Landauer’s bound. However, real-world computers operate many orders of magnitude above this limit. The main reason is that real systems are not ideal: they are constrained by their physical structure, they process information in finite time, and their components interact, leading to additional energy losses.

This project investigates how these realistic features influence the energy cost of computation. In particular, it focuses on three key aspects. First, it studies how interactions between components, such as bits or logic gates, lead to additional energy dissipation. Second, it analyzes how the mismatch between the expected and actual input of a computation increases energy consumption. Third, it examines how limitations in control, meaning that not all operations can be performed freely, affect energetic efficiency.

To address these questions, the project combines methods from physics, mathematics, and computer science. It develops theoretical models of simple computational systems and analyzes them using tools from modern non-equilibrium thermodynamics. These models are then extended to more realistic settings.

The goal of this project is to identify the fundamental mechanisms that determine the energetic costs of computation in real systems. By doing so, the project aims to provide general principles for designing more energy-efficient computing technologies in the future.

This project is funded by the Austrian Science Fund (FWF) 10.55776/PAT1151725.