IBM links two 120-qubit processors with four metre-scale superconducting cables

Researchers at IBM Research have joined two 120-qubit superconducting processors using four separate quantum links, each carried by a cable of roughly 80 cm. The links were fitted after both chips were already installed, and they transferred quantum states with an average accuracy of 99%. The work, led by Neereja M. Sundaresan and colleagues, is described in a preprint on arXiv and has not yet been peer reviewed.

Why modular designs need links

Many researchers expect that larger quantum computers will be built from several interconnected modules rather than a single enormous chip. In such a design, separate processors must exchange quantum information without degrading it. For superconducting hardware, that means building dependable connections about a metre long inside refrigerators cooled to just above absolute zero.

Until now, links between superconducting chips have typically been a single connection between small experimental devices. Those demonstrations showed that quantum states could be moved between chips and that entanglement could be generated across them. Scaling up requires something different: many closely packed connections, attached to processors already mounted in their cryogenic systems, with very low signal loss and the ability to survive repeated warming and cooling.

How the links work

The team connected two of IBM's Nighthawk R1 processors, each with 120 qubits. In every link, a qubit on one processor hands its information to a dedicated interface qubit. An adjustable coupler then passes that information into a superconducting aluminium cable, which delivers it to a matching set of components on the second chip.

To fit more cables into a small space, the researchers plugged each one into its chip from directly above rather than from the side. This arrangement allowed the four links to sit close together.

Performance across the links

The team tested both state transfer and logic operations involving qubits on the two processors.

• Quantum state transfer reached an average accuracy of 99% across the four links, with the best link achieving 99.29%.
• Two-qubit logic operations spanning both processors reached an average accuracy of 94.75%.
• The researchers generated entangled states shared between the processors, including 4-qubit and 6-qubit states.
• Two entangled pairs were created simultaneously through neighbouring links.
• All four connections stayed operational through three cycles of warming and cooling.

The thermal-cycling result is notable for practical engineering, since any system assembled from modules will need to be cooled, warmed and serviced many times.

Limitations and next steps

The accuracy of cross-processor logic operations, at under 95% on average, is lower than the state-transfer figures, and further gains would be needed for demanding computations. The source also notes that improvements are required before the approach could be commercially practical, particularly in how long quantum information survives. Three thermal cycles is also a modest test compared with the repeated cycling a deployed system would face.

Even so, the demonstration shows that several metre-scale links can be added to large, already installed processors and operate together. If the fidelity and coherence can be improved, this kind of architecture could allow quantum information to be stored and processed across many processors, making future machines easier to expand and maintain. The result will be of interest to groups working on quantum error correction and on distributed quantum architectures, where inter-module connectivity is a central constraint. Peer review and independent replication will be the next tests of the findings.

Source: Four superconducting cables link two large quantum processors into one (Phys.org, 31 October 2026).

Image: AI-generated illustration.