A chip-sized heat engine tests the path around quantum computers' million-cable problem

One chip-sized component that can flip between heat source and heat sink is now the center of a new result from Aalto University. It is a proof-of-concept cyclic quantum heat engine, the first one built on a superconducting circuit, and it was published in Nature Communications on 13 July 2026. The paper, led by first author Tuomas Uusnäkki and Academy Professor Mikko Möttönen, is incremental on its own. What it points at is the bottleneck that decides whether the field reaches large, useful quantum machines.

The practical problem is mundane, which is why it has been easy to overlook. A superconducting quantum computer is wired like a server rack drawn with a fine-tip pen. Each qubit needs dedicated microwave control lines that travel from a rack at room temperature down to the dilution cryostat, where the chip sits at millikelvin temperatures, or thousandths of a degree above absolute zero. Scale that wiring up to the size of machine the field actually wants and the cable count becomes its own engineering project.

Aalto's press release frames Finland's national target in those terms. The country's Quantum Technology Strategy calls for a 1,000-logical-qubit quantum computer by 2035. Möttönen estimates that target implies hundreds of thousands of physical qubits, because error correction has to spread its overhead across many physical devices to build one useful logical qubit. At the cabling density current stacks assume, the team argues, that means millions of microwave lines routed from millikelvin to room temperature, each cited at roughly €1,000, before counting the noise each line drags in. Wire counts at that scale become a cost and a heat problem at the same time.

The Aalto device is a step toward treating that problem on the chip rather than at the rack. The team built a flux-tunable transmon qubit, a common superconducting quantum bit, coupled to a resonator and to a voltage-tunable quantum-circuit refrigerator, a small on-chip component that can absorb energy from the qubit. By switching the refrigerator between heating and cooling modes, the researchers use a single component as both the hot and cold reservoir of a quantum Otto cycle, a four-stroke thermodynamic cycle adapted to a quantum system. Timed pulses on the qubit's flux line and the refrigerator's drive line execute the strokes. Single-shot readout tracks the qubit's energy populations through a four-component Gaussian mixture model. The device showed positive work output and finite efficiency across up to three measured cycles.

The novelty sits in the cycle, not the physics. Cyclic quantum heat engines have already been demonstrated in trapped ions, nuclear magnetic resonance, nitrogen-vacancy centers in diamond, cold rubidium, and cesium-rubidium collision systems. The Aalto paper's literature framing notes that the only prior experimental thermal-machine claim in superconducting circuits was non-cyclic. This is the first cyclic one in that platform. The distinction is what matters for the scaling story: an engine that runs on a real thermodynamic cycle, in hardware that is also a candidate for the kind of large-scale quantum computer Finland wants to build.

What it does not yet show is the autonomous on-chip controller the team wants this to grow into. Möttönen describes the result, in the institutional release, as a proof of concept on the way to heat engines that could eventually handle tasks such as qubit readout without sending control pulses up to room temperature. The paper reports a single-transmon device running an Otto cycle, with the dataset and analysis code deposited on Zenodo. Whether a similar device can drive a full quantum processor, or whether the same idea generalizes to many qubits, is not established here.

The work was done at OtaNano, Finland's national research infrastructure for nano- and quantum technology, with funding from the Research Council of Finland and the Finnish Cultural Foundation. The Qubit Report's re-report covers the same paper as discovery context but adds no new mechanism beyond the Aalto release and the Nature paper.

The first thing a 1,000-logical-qubit machine will need, before it needs another qubit, is a less tangled way to control the ones it already has. The Aalto result is a small, well-defined step in that direction. The next test is whether the same on-chip thermal trick scales beyond one transmon.