Light is the perfect carrier of quantum computing, the Netherlands-based quantum computing company Quix says. And now Quix says it has delivered the first-ever universal photonic quantum computer to customers. The new computer, named Carina, uses single photons as physical qubits, operates at room temperature and fits in standard data center racks. It offers eight input photonic qubits and four computational photonic qubits, which the company says it will scale quickly.

“What QuiX Quantum is showing with Carina and its measurement-based approach is that this path is not only tractable but navigable with integrated photonics,” says Gerard J. Milburn, a professor at the University of Queensland who was an early pioneer in photonic quantum technology. "It moves the conversation from whether photonic quantum computing can be universal to how quickly it can be scaled.”

Most quantum computers, of course, live in laboratories at very low temperatures, wrapped in cryogenic plumbing and tended by physicists. QuiX Quantum wants to put quantum computers into standard data centers at room temperature right next to the classical machines they are meant to work alongside.

As such, Carina is the world’s first universal photonic quantum computing architecture designed for deployment in customer environments rather than a specialized quantum facility, Quix says. The system was built as part of the Universal Photonic Quantum Computer project of the German Aerospace Center's Quantum Computing Initiative (DLR QCI), funded by the German Federal Ministry of Research, Technology and Space, and its core hardware has already been delivered to the DLR QCI.

A word doing some heavy lifting in this announcement, of course, is "universal."

Existing photonic quantum machines have been largely special-purpose: boson samplers and similar devices that can demonstrate quantum phenomena and maybe even quantum advantage on narrow tasks, but aren’t necessarily reprogrammable to run arbitrary algorithms. In contrast, Carina is designed to implement a universal gate set, meaning that in principle it can execute any gate-based quantum algorithm. QuiX says it’s built to run small-scale demonstrations of textbook algorithms such as Shor’s, Grover’s, Deutsch-Jozsa and quantum teleportation. That’s a proof of architecture, not a threat to global encryption … the company is careful to say Carina can’t break codes or solve commercially meaningful versions of those problems.

QuiX’s core quantum bet is architectural. Rather than build gates directly into the hardware, a notoriously hard thing to do with photons because they barely interact with each other, Carina uses a measurement-based approach. It prepares a large, highly entangled "cluster state" and drives the computation through a sequence of adaptive single-qubit measurements, with each result informing the next via fast feed-forward control. Gate-model algorithms are translated into these measurement-based operations through a compiler layer. The two models are mathematically interchangeable, so a circuit written for a conventional quantum computer can be mapped onto Carina.

Under the hood, the machine stitches together pieces QuiX has been announcing recently: on-chip photon generation at room temperature, high-speed switching and multiplexing, cluster-state generation, a Feed-Forward Control Unit that turns photon-detector signals into real-time control actions and a Photonic Assembly Control Unit that provides a standardized control layer for the photonic chips.

QuiX also says it recently demonstrated a production-ready form of "below threshold" error mitigation on a photonic quantum computer: a first for a European quantum company.

On the face of it, the specs don’t jump off the page. QuiX is downplaying the big number I always ask about: qubit count. Carina has eight input photonic qubits and four computational photonic qubits. The company frames this as a first-transistor moment. The point, in other words, is to show the architecture works and can be delivered, after which the work of scaling qubit count and quality begins.

QuiX’s room-temperature, discrete-variable photonic route meant to scale inside ordinary data centers is an explicit contrast with approaches that lean on continuous-variable photonics, cryogenics, proprietary facilities or quantum-as-a-service models. The goal, Quix says, is to let governments, enterprises and HPC operators run quantum infrastructure in their own environments now and build operational muscle around it before utility-scale machines arrive.

Of course, how soon QuiX can scale this is an open question. But the transition is important, moving photonic quantum from a can-this-be-done to when-will-we-get-more conversation.

I asked Quix for more information; here are answers from Stefan Hengesbach, CEO of QuiX Quantum.

John Koetsier: What can you compute that previous photonic systems can't?

Hengesbach: Carina is designed to move photonic quantum computing from special-purpose experiments toward a programmable, universal photonic platform. The focus is on universality through a measurement-based architecture at the hardware level rather than gate-based hardware. Carina executes gate-model algorithms through a compiler layer that translates quantum circuits into measurement-based photonic operations. Measurement-based and gate-based methods are interchangeable for quantum algorithms, e.g. algorithms in gate-based systems can be translated to measurement-based systems and vice versa.

Carina is built to run small-scale demonstrations of algorithms such as Shor's, Grover's, Deutsch-Jozsa and quantum teleportation. That doesn't mean Carina can break encryption or solve commercially relevant versions of those algorithms today. Instead, it demonstrates a deployable universal photonic architecture required to run this class of algorithms and provides the foundation for scaling toward larger, fault-tolerant systems. Current photonic systems are unable to deliver a universal computing platform that can be integrated in HPC/data center environments.

John Koetsier: Are you releasing the number of physical and/or logical qubits? If so, how many?

Hengesbach: Yes. Carina has eight input photonic qubits and four computational photonic qubits. At this stage, Quix is focused is on the architecture not the qubit count. Carina is intended to demonstrate that Quix can build and deliver a universal photonic quantum computing platform. Once that architecture is demonstrated, the next phase is scaling the number and quality of photonic qubits. We compare our approach philosophy to building the first transistor in the 1960s and then scaling from there.

John Koetsier: Any guidance on pricing?

Hengesbach: Carina is not an off-the-shelf product with a fixed list price. It is a customer-specific quantum computing platform that includes system design, integration, delivery and support. Pricing depends on the configuration, scope and deployment requirements. As a reference point, systems of this class are multi-million euro strategic infrastructure projects.

John Koetsier: Any comparisons or points of reference vis-à-vis PsiQuantum?

Hengesbach: There are different architectural paths within photonic quantum computing. Quix is taking a room-temperature, discrete-variable (DV) photonic approach focused on delivering universal photonic quantum computing. The system can be deployed into customer environments and scaled in standard data center settings. Other photonic approaches may involve different assumptions around continuous-variable (CV) photonics, cryogenic infrastructure, proprietary facilities or quantum-as-a-service models.

QuiX’s strategy is to build quantum computers that can ultimately scale within normal data center environments, rather than requiring Quix to own and operate a specialized quantum data center. Carina is designed for room-temperature operation, rack-based integration with classical computing and HPC workflows. The goal is to enable government, enterprise and HPC customers deploy and use quantum computing infrastructure in their own environments, including on-premise as well as offering their own cloud or quantum-as-a-service models.

Carina brings a universal photonic quantum computing architecture into a customer setting now, rather than building a large-scale independent facility in the future.

John Koetsier: Any comparisons or points of reference vis-à-vis Xanadu?

Hengesbach: Quix is pursuing a different route than all other photonic quantum companies: a room-temperature, discrete-variable photonic architecture built as a measurement-based universal photonic quantum computer. Carina is designed to show that this universal photonic system can be delivered into a customer environment, integrated with classical computing workflows and used as the basis for future scaling.

John Koetsier: Any comments on fault tolerance?

Hengesbach: Fault tolerance remains the long-term goal for the entire quantum computing industry. Quix recently released its Dedalo white paper, which outlines the company's architecture plan for scaling toward logical qubits. Carina is an important step on that path because it demonstrates a universal architecture that is a prerequisite for utility-scale quantum computing. The next phase is focused on improving photonic qubit quality and scaling the architecture in a way that can support logical qubits and fault-tolerant systems.

In addition, there are also quantum computing developers who believe that a measurement-based approach could be more effective than a gate-based approach when it comes to error-correction. Traditional error correction codes, such as surface codes, assume qubit locality—meaning a qubit can only interact with its immediate neighbors which you find in gate-based systems. This restriction limits the efficacy of error correction. In contrast, photonic qubits offer all-to-all connectivity, enabling more efficient error correction and reducing the hardware overhead required for scalability. Measurement-based also opens the possibility for more effective error correction schemes for in-code errors. By removing locality constraints, overall system size can be reduced while maintaining robust error correction.

Additional thoughts on scaling:

It should be noted that photonic quantum computing offers a unique scaling path via optical interconnects. Other modalities will ultimately need photonic links to connect modules while preserving quantum information. Quix's approach is native to photonics, which means the communication and interconnect layer is not an add-on; it is part of the architecture from the start.

John Koetsier: Thank you.