Cisco Wants Quantum Networks to Work Like Cloud Computing: Entanglement on Demand

On October 7, Cisco quietly released two of the most consequential pieces of quantum software announced this year. Through its Outshift innovation division, the networking giant introduced a Quantum Network Controller and version 0.2.0 of its Network-Aware Quantum Compiler, two research prototypes designed to turn entanglement, the fragile linked quantum state that distant machines share, into a resource an application can simply request. No hand-built device drivers, no link-by-link configuration, no physicist standing by the rack.

The idea is a direct import from cloud computing. Just as shared infrastructure turned CPU cycles from something you owned into something you scheduled, Cisco proposes that a quantum network should deliver entanglement as a service, with software handling who gets it, at what quality, in what order, and on what timing. The company calls the model Entanglement-as-a-Service, and it is the clearest statement yet of where Cisco thinks quantum networking is going: not bigger individual machines, but many machines behaving as one, run by software.

The problem with quantum networks today

Quantum networks have so far been built one point-to-point connection at a time. A handful of nodes connected by dedicated links can be run with hand-built software, which is painful but workable. The math breaks down fast: connecting 1,000 nodes point to point requires close to 500,000 dedicated links, since every pair needs its own connection. Nobody is hand-managing half a million quantum links.

Cisco's answer is a shared fabric. Earlier this year the company introduced its Universal Quantum Switch, a research prototype designed to route quantum information between devices that speak different physical languages, so that every node can reach every other node over one switching system instead of dedicated wires. Reported figures from that prototype described polarization-based switching at nanosecond speeds drawing under a watt. The new Controller is the operating layer above that switch: the software that decides, continuously, which routes get built, for whom, and when.

What the Controller actually does

The Controller exposes one interface per device category, sources, switches, detectors, and timing systems, with a hardware abstraction layer underneath. An application names the endpoints, the rate, the fidelity, and the timing it needs, and the system figures out how to physically produce the result. Hardware from multiple vendors plugs into the same category interface: Cisco names Qunnect of Brooklyn and Swabian Instruments as vendors whose sources, switches, and time taggers work through the abstraction.

Running a quantum network is stranger than running a classical one. A quantum state cannot be read without destroying the information it carries, so the Controller cannot inspect traffic the way conventional monitoring tools examine ordinary data. Instead it checks link quality statistically and applies predefined responses when performance drifts: tuning equipment, retrying, or escalating to a human when self-corrections fail. It reclaims hardware when each job ends, which matters in a shared network.

The ambition is to make entanglement a scheduled utility rather than a physics experiment: applications order a connection with a rate, a fidelity, and a deadline, and the network sorts out the photons.

What has been tested

The most concrete evidence comes from a February 2026 deployment, eight months before the Controller itself was introduced. Cisco software coordinated multi-node entanglement distribution and swapping across 17.6 kilometers, about 11 miles, of deployed commercial telecom fiber in New York City, using hardware from multiple vendors. Polarization fidelity, roughly how closely the delivered quantum states matched the intended ones, came in above 99 percent, at room temperature. Running at room temperature matters because much quantum hardware requires deep cryogenics; showing that the network coordination layer works with ordinary fiber and ambient-temperature optics is a meaningful step toward deployment outside the lab.

That run demonstrated the coordination approach, not the finished Controller. But it is the kind of evidence the field has been waiting for: entanglement swapping across separate fiber links, managed by software, with third-party hardware, outside a physics laboratory. Entanglement swapping is the core operation for extending entanglement across longer distances and more nodes, so showing it works over commercial fiber with above-99-percent fidelity is a real, if incremental, advance.

The compiler side of the stack

The Network-Aware Quantum Compiler v0.2.0 is the Controller's first native application, and it is designed to get no special access: it uses the same general-purpose interface any third-party compiler would use. A compiler translates a program into instructions hardware can execute; Cisco's version also decides how to divide a quantum program among separate processors and what network resources that division requires. It works out which nodes need entanglement, how much, and at what fidelity, then hands those requests to the Controller.

Version 0.2.0 integrates circuit compilation, qubit partitioning, and distributed quantum error correction against physical network parameters. It partitions programs with graph-cut strategies, schedules the Bell-pair generation needed for lattice surgery, the operation that stitches error-correcting codes across modules, and computes a Total Logical Error Rate, a single metric that lets architects compare network topologies and error-correction tradeoffs before any hardware is built. Cisco is offering a free 30-day demo of the compiler, a sign it wants outside developers to start building on the abstraction now.

What this is not, yet

Cisco is explicit that both systems are research prototypes, and the announcement does not establish that a large network of connected quantum processors is ready for commercial use. The field's central difficulty remains: single processors hold fewer qubits than useful workloads require, and linking them introduces a scheduling and timing problem that is genuinely harder than classical networking. Entanglement generation is probabilistic, qubits decohere while the network works, and operations demand timing precision measured in billionths of a second.

There is also a competitive question the announcement leaves open. Cisco's bet is that the winning position in quantum networking is the operating layer, the software that schedules and abstracts, rather than any single qubit technology. Other players are betting on hardware: faster interconnects, better memories, fault-tolerant processors. Cisco's own collaboration with Infleqtion on networked neutral-atom research has, as of October, defined an orchestration layer rather than produced an integrated system. The Controller is a proposed answer to a real problem, not a deployed product, and the next useful disclosures will be end-to-end workloads across separate nodes, not more architecture diagrams.

In perspective

The pattern here is familiar to anyone who watched classical networking mature. First came the physical links, then the protocols, then software-defined networking, which separated the control plane from the hardware and let operators program networks instead of configuring boxes. Cisco, which rode that transition to dominance in classical networking, is attempting the same move in quantum: own the control plane, the abstraction layers, and the unified metrics, and let the qubit technologies compete underneath.

That strategy looks increasingly plausible precisely because the hardware side is accelerating. Photonic interconnects are hitting kilohertz rates, error-correction benchmarks are being standardized, and funding for fault-tolerant machines is flowing. Cisco's argument is that distributed quantum computing will need an operating system before it needs a miracle, and that the entanglement scheduling problem is one software can solve today. Whether the industry adopts its abstraction or builds its own, the era of hand-managed quantum links is clearly ending.