Cisco Develops Control Layers for Scalable Quantum Networks

Building a Networked Quantum Architecture

Cisco Networks, through its Outshift research and incubator unit, has spent several years developing the network architecture needed for quantum computing. The company’s view is that scaling quantum environments will require more than building larger systems with greater numbers of qubits. Quantum computers will also need to connect through networks, creating distributed and cloud-like architectures similar to those used in classical computing.

In other words, quantum computing will need to scale out, not only scale up.

Early last year, Outshift was tasked with developing Cisco’s quantum networking stack. Later in 2025, the group unveiled a prototype quantum network entangling chip that generates entangled pairs of photons across large distances. Entanglement cannot be produced with classical bits of information and is a central part of quantum networking.

In April, Cisco introduced a working prototype of its Universal Quantum Switch. The switch is designed to connect quantum sensors and systems from different vendors and based on different modalities, including superconducting systems, photonics, neutral atoms, and trapped ions.

Cisco is now developing the tools needed to scale the network itself. At Cisco’s Quantum Summit this month, the company announced the Quantum Network Controller, another research prototype. It provides a control layer intended to replace manual configuration with automation capable of handling the machine-speed, any-to-any connections enabled by the Universal Quantum Switch.

“Connecting 1,000 quantum nodes point to point takes close to 500,000 dedicated links,” wrote Vijoy Pandey, senior vice president of Outshift.

Automating Quantum Network Operations

According to Ramana Kompella, head of Cisco Research, much of the work in today’s early quantum networks is still performed manually. That approach will not scale as demand increases.

“Where we are today, if you look at the quantum networks, they're very point-to-point, manually stitched together and configured,” Kompella said. “The moment you actually have application scale requests coming in at machine speeds, essentially saying, ‘I want to be able to entangle this node and this node,’ that's the power of a network. Any nodes would want to properly exchange entanglements, so if you have these requests coming in at machine speeds, the human way of configuring and reconfiguring the fabric is not going to scale. That's why you need a control plane that can scale and operate at machine speeds, which is where the controller comes into play.”

Classical systems already use controllers to automate operations, manage resources, expose APIs, and perform other tasks. However, those systems are generally deterministic: execution timing, task scheduling, and other steps are predictable and repeatable.

Quantum systems are more complex and probabilistic, according to Reza Nejabati, head of quantum research and Quantum Labs at Cisco. Qubits have short-lived coherence times, while factors such as loss, fidelity, synchronization, and overall system performance can determine whether an operation succeeds. Cisco’s goal is to shield application developers from much of this complexity.

The controller acts as an abstraction layer based on what Pandey described as an Entanglement-as-a-Service model. It exposes APIs for components including switches, detectors, and timing systems. An application can request entanglement by specifying the endpoint, fidelity, rate, and timing requirements. The controller then schedules the connection and works to maintain the required link quality. It can apply predefined tuning and retry or reinitialize a job if performance declines.

Connecting Different Quantum Modalities

The controller is also intended to allow quantum systems based on different modalities to connect to the network. Cisco, like IBM and other companies, uses superconducting technology, but Kompella said future networks will include components from many vendors.

“We would love to think that everything will be built by Cisco, but the reality is we will work with many different components coming in from many different vendors,” Kompella said. “It's also the goal of this to try and hide the complexity associated with onboarding many different types of devices and shield all this complexity to the applications. That's really the missing piece without which you cannot actually scalably operate your quantum networks in the future.”

Cisco is also developing a Network-Aware Quantum Compiler. Applications will need to integrate with and communicate through the controller as distributed quantum computing develops.

“The first application that we tackle is the distributed quantum computing compiler, which effectively takes a large circuit, breaks it up into smaller pieces, and drives the network with the help of this controller in distributing entanglements at the right time so that your partial computed state is actually moving from processor A to processor B,” Kompella said. “Together, these are the missing pieces.”

Pandey described the controller and compiler as complementary. The compiler determines the computation required by an application, including the amount of entanglement, the nodes involved, and the required fidelity. The controller then translates those requirements into network operations.

Emerging Applications and Remaining Challenges

Quantum networking is attracting increasing attention. Some estimates place the market’s size at approximately $2 billion, with vendors including Nu Quantum, Photonics Inc., Aliro Technologies, and QphoX. Cisco also partnered with Qunnect earlier this year in a demonstration of scalable quantum entanglement swapping across 17.6 kilometers of New York City telecommunications fiber.

Kompella said quantum networking could support classical computing applications as well as quantum workloads. One example is decision coordination for high-speed trading in financial services, where a system such as Cisco’s Quantum Sync could enable rapid communication between geographically separated nodes.

Another possible application is quantum eavesdropper detection. This approach combines quantum-entangled photons with classical photons on the same optical fiber. If an eavesdropper interferes with the signal, the entanglements are disturbed, making the intrusion detectable.

Quantum networks remain at an early stage, but Kompella said interconnected quantum systems could emerge within three to five years. Significant technical challenges remain, including transduction. This process converts quantum signals and information from one physical domain or frequency to another while preserving properties such as superposition and entanglement. Kompella described it as the point “where your computational qubits meet the communication qubits.”

Quantum memory is another challenge. It involves storing data as qubits, which are fragile and can lose their quantum properties when disrupted by light, vibration, or other forms of noise.

“Quantum memories are actually extremely helpful in increasing your fidelities and timing and coordinating between many different quantum nodes,” Kompella said. “That's an area where quantum memories is still nascent and you probably need to see some more advances in realizing some of these cutting-edge applications.”