Can quantum technology make classical computers use less energy?

Quantum coherence is usually associated with quantum computing. A new theoretical proposal by Daniel Loss, RDIA Chair Professor at KFUPM in Saudi Arabia and founding co-director (now honorary director) of NCCR SPIN, asks whether it could also make classical computing more energy efficient.

Loss discusses the idea in a new episode of The New Quantum Era, hosted by Sebastian Hassinger.

Reversible computing is not new. Its foundations were developed several decades ago by Rolf Landauer, Charles Bennett, Edward Fredkin and Tommaso Toffoli. What has changed is the technology available to implement it. Recent progress in controlling and moving spins in semiconductor quantum dots brings the idea closer to an experimental test.

Conventional processors generate heat as they perform calculations. Part of this heat comes from irreversible logic operations, which overwrite or erase information. According to Landauer’s principle, erasing a bit has a minimum energy cost. Reversible computing preserves this information. Each operation contains enough information to recover its input from its output. Temporary results can be “uncomputed” instead of erased, reducing the energy dissipated as heat.

Loss proposes implementing this form of classical computation using coherent spin dynamics in germanium-silicon quantum dots. The hardware is quantum, but the information remains classical. Each spin represents a 0 or a 1 at the beginning and end of an operation. Quantum coherence controls the internal evolution of the gate, without any algorithmic use of superposition. Since the information remains in classical states between operations, this may also reduce the error-correction requirements compared with quantum computation.

The central component is an iToffoli gate. A target spin moves between quantum dots, while two neighboring spins control its evolution. The operation is driven by DC voltage pulses rather than radio-frequency signals. A Toffoli gate is reversible because its outputs contain enough information to reconstruct its inputs. Shuttling also allows the spin to carry information between quantum dots without being measured. Logic and data movement can therefore remain reversible until the final readout.

The energy estimate is striking. A spin-based Toffoli gate operating near 4 kelvin could use about 100,000 times less energy than its room-temperature CMOS equivalent, even when the estimated refrigeration overhead is included. The paper also identifies parts of the full energy budget that still need to be measured.

Loss suggests that AI inference could be an early application. Some inference workloads are highly parallel and can tolerate a limited level of error. They would not require new quantum algorithms because the calculations remain classical. The same quantum-dot array could support classical reversible logic or quantum algorithms, depending on how it is operated.

The next step is to test the proposal experimentally. Initial experiments could demonstrate a three-spin iToffoli truth table, measure the energy used by the gate and implement a five-dot reversible adder.

Listen to the episode: Classical Reversible Computing on Quantum Dots with Daniel Loss

Read the paper: Classical Reversible Computation by Quantum Coherence