IBM × Cisco × SQMS Quantum Networking Units (QNU) plan

IBM × Cisco × SQMS Quantum Networking Units (QNU) plan

A roadmap announcement, not an operating network. IBM, Cisco, and Fermilab's Superconducting Quantum Materials and Systems Center (SQMS) announced on 20 November 2025 a joint plan to network superconducting quantum processors across multiple cryogenic systems over optical fibre, with a proof-of-concept demonstration targeted for end-of-2030 and a production-class system in the early 2030s.

Operator
IBM (processors + QNUs) · Cisco (networking infrastructure) · Fermilab Superconducting Quantum Materials and Systems Center, SQMS (materials research)
Location
Multi-site programme (IBM Yorktown Heights, Cisco labs, Fermilab Batavia)
Year
Announced 20 November 2025; PoC target end of 2030; production-class system early 2030s
Technology
Superconducting qubits (IBM); microwave-to-optical transduction for inter-cryostat links; QNU network interface; optical-fibre entanglement distribution between fridges
Scale
Planned target — tens to hundreds of thousands of logical qubits distributed across multiple cryostats
Status
Roadmap announcement, no operating system to date
Commercial model
Vendor R&D programme; eventual integration into IBM Quantum cloud service is implied but not committed

What it is

IBM superconducting processors live inside dilution refrigerators held at roughly 10–20 mK. Scaling one processor past the tens-of-thousands-of-qubits range runs into the wiring, cooling power, and footprint limits of a single fridge. The QNU plan replaces that scaling path with a networked one: multiple fridges, each holding a Starling-class processor, linked by optical fibre and stitched together at the algorithm level into a single distributed computation. IBM × Cisco × Fermilab/SQMS Quantum Networking Units, 20 Nov 2025

Three components are named as the work to be done. The first is the microwave-to-optical transducer: superconducting qubits operate at 5–10 GHz microwave frequencies, which cannot travel between buildings over optical fibre. A transducer converts the microwave excitation into a telecom-band photon (and back) at single-quantum fidelity. The second is the Quantum Networking Unit (QNU) itself — IBM-developed network interface hardware at each processor node that handles the protocol layer above the transducer. The third is the inter-system entanglement-distribution protocol that delivers Bell pairs between cryostats so that gate teleportation, state teleportation, and distributed-algorithm primitives have the shared entanglement they need.

The division of labour follows the existing vendor strengths. IBM owns the processors and the QNU interface. Cisco owns the networking infrastructure outside the fridge — the optical-fibre plant, timing, and control plane between sites. Fermilab's SQMS centre contributes the underlying materials research that the transducers and the high-coherence niobium cavities rely on.

Verified claims

Things to note

  • The announcement describes a roadmap. No operational link between two IBM fridges has been demonstrated as of this writing; no fidelity or throughput figures have been published. The announcement sets out a statement of intent and a division of labour between the partners.
  • The three named components are active R&D areas. The press release identifies microwave-to-optical transducers, the QNU interface, and the inter-system entanglement-distribution protocol as items to be developed during the programme rather than off-the-shelf parts.
  • Microwave-to-optical transduction at single-quantum fidelity remains a leading technical challenge. Reported laboratory devices currently sit below the conversion efficiency and added-noise targets that an inter-cryostat link would require. See the transduction subject page for the current state of the art.
  • "Tens to hundreds of thousands of qubits" is a target figure. The figure does not separately quote physical-vs-logical qubit counts; the underlying mix affects the engineering by orders of magnitude. Detailed assessment will be possible once the proof-of-concept reports specifics.
  • The "early 2030s" timeline aligns with IBM's broader fault-tolerance roadmap — Starling-class processors are scheduled into the same window. Inter-system networking sits on top of that fault-tolerance roadmap: a transducer that is not required for a single-fridge fault-tolerant machine becomes part of the critical path once the architecture is distributed.
  • Distinct from the Cisco × Qunnect Brooklyn demonstration. The Brooklyn entry on this site documents a metro-scale entanglement-swap over 17.6 km of deployed New York City fibre using warm-rubidium photon-pair sources — a different physical platform, scale, and measurement context. The QNU plan is a forward-looking superconducting-qubit programme. The two are separate efforts.