Photonic Quantum Computing: How Light-Based Qubits Work and Why They Could Beat Superconductors

Why Build a Quantum Computer from Light?

Every major quantum computing approach carries a fundamental engineering burden. Superconducting qubits need dilution refrigerators that operate at 15 millikelvin. Trapped ions need ultra-high vacuum chambers and precision laser systems. Silicon spin qubits need millikelvin cooling and nanometer-scale fabrication. Each approach trades one set of engineering challenges for another.

Photonic quantum computing uses photons (light particles) as qubits. Because photons barely interact with their environment, they maintain coherence at room temperature without refrigeration. The core challenges are probabilistic gate operations and photon loss. PsiQuantum, Xanadu, and QuiX Quantum are the leading developers targeting scalable photonic systems by 2027 to 2030.

Photonic quantum computing offers a distinct trade-off. Photons barely interact with their environment, which means they maintain quantum coherence at room temperature without any cooling. They can travel through fiber optic cables with minimal loss. And they can be produced using existing semiconductor manufacturing equipment. These properties make photonic systems attractive for large-scale quantum computing, but they come with their own hard problems: photons are difficult to make interact with each other, and many photonic gate operations are inherently probabilistic.

First Principles: What Is a Photonic Qubit?

A classical bit is either 0 or 1. A photonic qubit encodes quantum information in a physical property of a single photon. The most common encodings are:

• Polarization encoding: Horizontal polarization = 0, vertical polarization = 1. A superposition of polarizations encodes a qubit in superposition.
• Path encoding: A photon traveling down one waveguide = 0, the other waveguide = 1. A 50/50 beam splitter puts the photon in superposition of both paths simultaneously.
• Time-bin encoding: A photon arriving in one time window = 0, a later window = 1. This is robust for fiber transmission over long distances.

All three encodings support quantum superposition and, through entanglement, multi-qubit quantum states. The choice of encoding depends on the application and the physical platform.

How Linear Optical Quantum Computing Works

Linear optical quantum computing (LOQC) is the dominant approach in photonic quantum computing. "Linear" refers to the fact that the optical components, such as beam splitters and phase shifters, perform linear transformations on the quantum states of photons. These components are passive: they do not amplify or add energy to the photons passing through them.

Beam Splitters: The Basic Gate

A beam splitter is the photonic equivalent of a quantum gate. A 50/50 beam splitter takes an incoming photon and puts it into an equal superposition of the two output modes. For a photon in mode 0 (one input port), the output state is (|0> + |1>) / sqrt(2): a perfect superposition of the two output paths. This is equivalent to a Hadamard gate in qubit notation. By choosing different splitting ratios and adding phase shifters, any single-qubit rotation can be implemented using a combination of beam splitters and phase shifters.

The Problem with Two-Qubit Gates

Single-qubit operations on photons are nearly perfect. The hard problem is two-qubit gates. In superconducting systems, two qubits interact through a physical coupling on the chip. Photons do not naturally interact with each other; they pass through each other without any interaction. Producing a deterministic photon-photon interaction requires a medium that is both highly nonlinear (to mediate the interaction) and quantum (to preserve coherence). No such medium exists at the quality level needed for large-scale computing.

The solution, proposed by Knill, Laflamme, and Milburn (KLM) in 2001 and the basis of most photonic quantum computing approaches, is to use measurement-induced nonlinearity. By measuring ancilla photons (extra photons added to the circuit) and performing feed-forward operations (adjusting the circuit based on measurement outcomes), it is possible to implement effective two-qubit gates. The catch: these gates only succeed with some probability, typically 50% or less per attempt. Failed gates produce known error states that can be detected and corrected, but the probabilistic nature means circuits must be designed to retry gate operations or use redundant resources.

Key Players in Photonic Quantum Computing

PsiQuantum: The 1 Million Qubit Roadmap

PsiQuantum, founded in 2016 and headquartered in Palo Alto, is the highest-profile photonic quantum computing company. The company's strategy centers on silicon photonics: building photonic circuits on silicon wafers using GlobalFoundries' existing semiconductor fabrication infrastructure. This is a deliberate bet on manufacturability over near-term performance.

PsiQuantum's approach is fault-tolerant from the start. Rather than building NISQ-era photonic devices, the company is designing for a fault-tolerant photonic quantum computer with roughly 1 million physical qubits. The strategy uses the photonic threshold theorem: if photon loss and gate error rates are below specific thresholds, quantum error correction codes can suppress errors exponentially. PsiQuantum has published that its silicon photonics platform approaches those thresholds in laboratory settings.

PsiQuantum has secured over $700 million in funding and partnerships with GlobalFoundries and government agencies in the US, UK, and Australia. The company's target is a fault-tolerant system in the late 2020s, though specific timelines have shifted. A 2027 milestone for a functional error-corrected photonic system has been mentioned in company communications, though this refers to intermediate milestones, not a full million-qubit machine.

Xanadu: Gaussian Boson Sampling and the Cloud

Xanadu, based in Toronto, takes a different approach. Rather than targeting fault-tolerant general-purpose quantum computing from the start, Xanadu has built and deployed near-term photonic quantum processors using Gaussian boson sampling. Their Borealis processor, announced in 2022, demonstrated quantum advantage on a Gaussian boson sampling task with 216 programmable squeezed light sources.

Gaussian boson sampling is not a general-purpose quantum computation. It is a specific sampling task that is believed to be hard for classical computers. However, it has potential applications in molecular vibronic spectra simulation, graph theory, and some optimization problems. Xanadu makes Borealis and earlier processors (X8, X12, X24) available via their Strawberry Fields SDK and their cloud platform.

Xanadu's longer-term roadmap targets fault-tolerant photonic computing using their GKP (Gottesman-Kitaev-Preskill) qubit encoding, which encodes a qubit in the quantum state of a harmonic oscillator (in this case, a photonic mode with many photons). GKP qubits offer a path to high-threshold error correction in photonic systems.

QuiX Quantum: European Leadership

QuiX Quantum, based in Enschede, Netherlands, is the leading European photonic quantum computing company. Their platform uses silicon nitride waveguide technology rather than silicon, which offers lower optical losses in the visible and near-infrared wavelengths. QuiX's 20-mode photonic processor, released in 2021, was the largest publicly demonstrated photonic processor in Europe at the time.

QuiX focuses on the photonic processor as an accelerator for specific applications: boson sampling, quantum simulation, and quantum-enhanced sensing. Their 2024 and 2025 systems extended to larger mode counts with improved loss characteristics. The company targets enterprise and research customers in quantum chemistry and machine learning acceleration.

Comparison: Photonic vs Superconducting vs Trapped Ion

Photon Loss: The Core Technical Challenge

Photon loss is the photonic equivalent of qubit decoherence in superconducting systems. When a photon is lost, the qubit information it carried is destroyed. Unlike a phase error, which can in principle be corrected by quantum error correction, photon loss is an erasure error: the qubit is simply gone.

Photon loss happens at every optical component. Beam splitters, waveguide bends, fiber connections, and photon detectors all have finite efficiency. State-of-the-art superconducting nanowire single-photon detectors (SNSPDs) achieve detection efficiencies above 98%, meaning 2% of photons are lost at detection alone. Waveguide losses in silicon photonics are measured in decibels per centimeter; a 10 cm path through a silicon waveguide might lose 10 to 30% of photons depending on fabrication quality. Silicon nitride waveguides, used by QuiX Quantum, offer lower propagation losses (below 0.1 dB/cm vs 2-3 dB/cm for silicon at telecom wavelengths).

For a photonic quantum computer with long circuit depths, photon loss is catastrophic without error correction. This is why PsiQuantum and Xanadu's fault-tolerant strategies are not optional features: they are prerequisites for useful photonic quantum computing beyond a few dozen photons.

Photonic Quantum Networks: A Unique Advantage

One area where photonic systems have a clear and near-term advantage over superconducting and ion trap systems is quantum networking. Photons are the natural carrier of quantum information over long distances. Superconducting qubits operate at millikelvin temperatures and cannot be directly connected to room-temperature fiber optics without losing their quantum state. Sending a qubit between two superconducting quantum computers requires converting the microwave-frequency qubit to an optical photon (using a "quantum transducer"), a process that is extremely lossy and technically immature as of 2026.

Photonic systems have no such conversion problem. Photonic qubits at telecom wavelengths (1310 nm or 1550 nm) travel through standard optical fiber with losses of 0.2 to 0.3 dB per kilometer, limited by Rayleigh scattering. This makes photonic systems the leading candidate for quantum key distribution networks and the future quantum internet. Companies including Toshiba, ID Quantique, and QuantumCTek have already deployed photonic quantum key distribution networks over distances of 100 to 1,000 km using satellite and fiber links.

The Crypto Threat Angle: Photonic Systems and Q-Day

For the blockchain and cryptography communities, the question is not just whether photonic quantum computers work. It is how quickly they could scale to the point of being cryptographically relevant. For the same reasons that photonic systems are harder to scale than superconducting systems in the near term (probabilistic gates, photon loss), they are also less likely to be the first platform to achieve cryptographically relevant fault-tolerant quantum computing.

IBM and Google's superconducting platforms, with their mature error correction demonstrations and detailed public roadmaps, remain the most likely first platforms to cross the fault-tolerant threshold. Photonic systems, particularly PsiQuantum's silicon photonics approach, represent a viable path to massive scale (millions of qubits) but on a longer timeline. The photonic approach may ultimately enable larger-scale quantum computers than superconducting technology, but the superconducting systems will almost certainly get to the cryptographic threat level first.

This does not make photonic quantum computing irrelevant for security planning. A world where superconducting systems can break RSA-2048 is a world where photonic systems are also rapidly approaching that capability. The heterogeneity of the quantum computing landscape means that cryptographic systems need to be secure against multiple hardware approaches, not just the leading one. For a broader view of how the quantum computing landscape threatens existing encryption, see our analysis of how many qubits it takes to break Bitcoin, and our comparison of qubit types across all major platforms.

For a deeper look at which quantum computing companies are closest to a cryptographic threat and how to evaluate their progress, see our Q-Day company ranking.