Xanadu and photonic quantum computing

Two machines, two jobs

Xanadu quantum computing takes the same broad bet as PsiQuantum, that light is the most practical medium for building a large machine, but the Canadian company has split its work into two clearly separate systems. The first, Borealis, is a specialist machine with 216 qubits built for Gaussian boson sampling, a narrow task used to demonstrate quantum advantage, the point at which a quantum device does something no ordinary computer can match in reasonable time. The second, Aurora, is a smaller universal, gate-based system of 12 qubits aimed at the long road towards fault tolerance. The two architectures serve two different purposes, and understanding why is a useful way into how the field actually works.

Borealis and Aurora

Borealis and Aurora answer different questions. A boson-sampling machine like Borealis is not a general-purpose computer; it is tuned to one mathematical problem that happens to be extremely hard for classical hardware, which makes it a good instrument for proving that a quantum device can outrun a conventional one. Aurora, by contrast, is universal and gate-based, meaning it is built to run arbitrary quantum programmes in the way a full quantum computer eventually must. Its 12 qubits are modest by design, chosen to develop the architecture and error correction that a fault-tolerant photonic machine will require rather than to post a large headline number. Keeping the two projects distinct lets Xanadu quantum research chase a demonstration result and a long-term goal at once, without pretending that one machine can do both.

The manufacturing case

The manufacturing argument is where Xanadu and PsiQuantum align most closely. Xanadu builds its systems using standard photonic integrated circuit manufacturing, the established processes for making optical chips. The case both companies make is that photonics is the most manufacturing-compatible of the quantum modalities, because it can lean on techniques the electronics and telecoms industries have already refined over decades. If scaling to very large qubit counts is ultimately a fabrication problem, then starting from proven chipmaking methods is a genuine advantage. Optical chips can, in principle, be produced on the same kind of production lines that already make components for data centres and telecoms, which is why manufacturability sits at the heart of the photonic case.

The unsolved problems

The photonic route still has hard, unsolved problems, and they should be stated plainly. Reliable single-photon sources, which must produce exactly one photon on demand, remain difficult. So do fast, efficient detectors, and quantum memory, the ability to hold a photonic state still until it is needed, given that photons naturally travel at the speed of light. These are the open questions that stand between the photonic approach and a working fault-tolerant machine, and progress on Xanadu quantum systems will be judged partly on how far they close them.

Why two tracks, and what to watch

That two-track structure shows how a serious quantum company hedges. Borealis provides evidence, today, that Xanadu can build hardware capable of quantum advantage on a chosen problem. Aurora provides a platform for the slower, harder work of universal fault-tolerant computing. One demonstrates capability; the other builds towards usefulness. Together they let Xanadu quantum development show results now while investing in the architecture that decides whether photonics can go the distance.

Where Xanadu sits in the wider field is as photonics’ second major standard-bearer alongside PsiQuantum, and as evidence that the modality can support both advantage demonstrations and gate-based research at once. What to watch is progress on the underlying components, the single-photon sources, detectors and memory that will decide whether photonic machines scale, and whether Aurora’s universal architecture matures into something that can host real error correction. Our quantum explainer hub sets Xanadu against the competing designs, where its photonic approach shares both the promise and the open problems that PsiQuantum’s does.