Photonic Quantum Computing

Unlike superconducting qubits that require cooling to 15 millikelvin—colder than outer space—photonic qubits operate at room temperature. This eliminates massive dilution refrigerators that cost millions of dollars, dramatically reducing system complexity and enabling datacenter-scale deployment. While superconducting systems from Google and IBM require complex cryogenic infrastructure, photonic systems need only standard electronics and optics.

The theoretical foundation for photonic quantum computing was established in 2001 when Knill, Laflamme, and Milburn proved that universal quantum computing is possible using only linear optical elements: beam splitters, phase shifters, and single-photon detectors. This KLM theorem showed that while individual gates are probabilistic (succeeding only a fraction of attempts), heralding signals indicate success or failure, enabling fault-tolerant architectures through repeat attempts.

The Temperature Advantage

Superconducting quantum computers from Google and IBM operate at 15 millikelvin—approximately 180× colder than the 2.7 K cosmic microwave background. Each system requires complex dilution refrigerators with substantial infrastructure costs. The cooling requirements limit system size and introduce thermal management challenges as qubit counts grow.

Photonic systems operate at room temperature in the photonic circuitry, with cryogenic single-photon detectors (typically ~4 K for SNSPDs). Even this modest cooling requirement is orders of magnitude simpler than millikelvin refrigeration. PsiQuantum's architecture places detectors in a small 4K stage while the rest of the system—including millions of optical components—operates at ambient temperature.

Quantum Networking: The Killer Advantage

Photons are nature's information carriers—already used in fiber optic networks worldwide. Photonic qubits travel through standard telecom fiber (1550nm wavelength), enabling quantum networks without converting between qubit types. While superconducting and trapped ion systems require transducers to convert between stationary and flying qubits, photonic systems use identical qubits for both.

This enables modular quantum computing: connect multiple quantum processors through optical fiber to create larger systems. PsiQuantum demonstrated 99.72% chip-to-chip quantum fidelity—sufficient for fault-tolerant distributed computation. The same technology enables the quantum internet, quantum key distribution, and distributed quantum computing across datacenters.

Manufacturing at Semiconductor Scale

Silicon photonics leverages 50 years of semiconductor manufacturing expertise. PsiQuantum's partnership with GlobalFoundries produces quantum chips on standard 300mm silicon wafers using CMOS-compatible processes. This means quantum components can be manufactured at the same fabs that produce conventional microprocessors, with similar yields and costs at scale.

The Omega chipset represents PsiQuantum's breakthrough in manufacturable quantum computing. Integrating all critical components on a single platform—silicon nitride waveguides, barium titanate switches, superconducting single-photon detectors, and high-performance optical couplers—the Omega chips are manufactured using CMOS-compatible processes at GlobalFoundries fabs in Albany, NY and Dresden, Germany. This partnership marks the first time quantum chips are produced at true semiconductor scale.

Each Omega chip requires a 25-layer stack with over 500 process steps. Despite this complexity, manufacturing yields reportedly match standard semiconductors—remarkable for quantum-grade components. The Nature 2025 paper demonstrated extraordinary fidelity metrics: 99.999% interferometer precision, 99.98% state preparation and measurement accuracy, 99.72% chip-to-chip quantum fidelity, and 99.5% two-photon quantum interference visibility.

Fusion-Based Quantum Computing

PsiQuantum employs fusion-based quantum computing, where entanglement is created through photon pair fusion rather than direct two-qubit gates. When two photons are measured together, they can become entangled with remaining photons in the system. This fusion approach offers inherent loss tolerance—when fusion fails (detected through heralding), the system compensates through repeat attempts without corrupting the quantum state.

Fusion success rates of approximately 50% are sufficient for fault tolerance when combined with appropriate error correction codes. The 2023 Nature Communications paper established fusion-based quantum computing as a viable pathway, demonstrating that realistic component imperfections can be corrected through redundancy and measurement-based error correction.

GlobalFoundries Partnership

GlobalFoundries, which originated from AMD's manufacturing operations in 2009, operates facilities in New York, Dresden (Germany), and Singapore. The company produces chips for AMD, Qualcomm, and other major clients using advanced semiconductor processes. PsiQuantum's partnership leverages this infrastructure for quantum chip production at unprecedented scale.

The partnership began in 2021 with qualification of quantum-specific processes. By 2024, GlobalFoundries was producing Omega chips at volume, demonstrating feasibility of quantum manufacturing at semiconductor industry standards. The collaboration includes co-development of specialized processes for superconducting detectors and electro-optic modulators.

Datacenter-Scale Facilities

Construction began in 2025 on PsiQuantum's first datacenter-scale quantum computing facilities: a $940M AUD facility in Brisbane, Australia (supported by federal and state governments), and a facility in Illinois backed by state investment. These facilities will house thousands of interconnected chips, millions of components, and the infrastructure for fault-tolerant quantum computing.

PsiQuantum: Manufacturing First

Founded in 2016 by Terry Rudolph (Imperial College), Jeremy O'Brien (Bristol), Mark Thompson, and Pete Shadbolt, PsiQuantum made an early strategic decision: design for mass manufacturing at commercial semiconductor fabs rather than lab prototypes. This bet on manufacturability has paid off with the GlobalFoundries partnership producing Omega chips at semiconductor industry scale.

The company has raised over $700M in venture funding from BlackRock, Microsoft's M12, Baillie Gifford, and other major investors. Government partnerships include $940M AUD from Australia for a Brisbane datacenter facility (announced 2024), and state funding from Illinois for a US facility. Commercial operation is targeted for 2027 with fault-tolerant systems expected by 2029.

PsiQuantum's technical approach uses fusion-based quantum computing, where photon pair measurements create entanglement. The fusion gates are inherently loss-tolerant—when fusion fails, the failure is detected without corrupting quantum information. Combined with topological error correction codes, this provides a pathway to fault-tolerant quantum computation.

Xanadu: Quantum Advantage Achieved

Founded in 2016 in Toronto, Xanadu has pursued continuous-variable (CV) quantum computing using squeezed light rather than single photons. In June 2022, their Borealis system became the first photonic quantum computer to demonstrate computational advantage—solving Gaussian Boson Sampling in 36 microseconds that would take classical supercomputers 9,000 years to simulate.

The Borealis system uses 216 squeezed-state qubits with programmable linear optical networks. Unlike PsiQuantum's fusion approach, Xanadu encodes quantum information in the continuous properties of light (amplitude and phase quadratures). For error correction, they're developing GKP (Gottesman-Kitaev-Preskill) states—grid codes that encode discrete information in continuous variables.

In 2025, Xanadu announced on-chip generation of GKP states, a significant step toward fault-tolerant CV quantum computing. The company also announced plans to pursue a $3B IPO, which would make it the first publicly traded pure-play photonic quantum computing company (per Xanadu press release).

Quandela: European Champion

Spun out from CNRS and Université Paris-Saclay in 2017, Quandela focuses on discrete-variable photonic computing using quantum dot single-photon sources. Their MosaiQ system delivers photons with over 97% indistinguishability—critical for high-fidelity quantum interference.

Quandela is positioning as the European sovereign quantum computing platform, with installations at major research centers and commercial partnerships across the EU. The company provides cloud access through their Quandela Cloud platform, democratizing access to photonic quantum computing.

Photonic Inc: Spin-Photon Hybrid

Based in Vancouver and operating in stealth until 2022, Photonic Inc. takes a unique approach: using silicon T-centers (color centers) that combine long-lived spin qubits with native optical interfaces. Each T-center contains a spin qubit for computation with direct optical readout for networking—eliminating the need for transducers between stationary and flying qubits.

The quantum internet will connect quantum computers across distances, enabling distributed quantum computing, blind quantum computation, and secure communication through quantum key distribution. Optical photons are the leading carriers for long-distance quantum communication—they travel at the speed of light through existing telecom fiber infrastructure with low decoherence.

While superconducting and trapped ion systems require transducers to convert between stationary (computation) qubits and flying (communication) qubits, photonic systems use the same qubit type for both operations. This eliminates conversion losses, simplifies system architecture, and enables seamless scaling through modular chip-to-chip connections.

Chip-to-Chip Quantum Interconnects

PsiQuantum's 2025 Nature paper demonstrated 99.72% fidelity for quantum state transfer between chips through optical fiber. This enables modular architectures where multiple chips work together as a unified quantum computer. Unlike electrical interconnects, optical fiber can transmit quantum states with low decoherence, though loss grows exponentially with distance (hence the need for quantum repeaters).

The chip-to-chip interconnect includes: (1) on-chip generation of entangled photon pairs, (2) low-loss optical coupling to fiber, (3) transmission through standard telecom fiber, (4) high-efficiency detection at the receiving chip. Each step has been optimized to minimize loss while maintaining quantum coherence.

Telecom Wavelength Compatibility

Photonic systems operate at 1550nm wavelength—the sweet spot for telecom fiber networks where attenuation is minimized (~0.2 dB/km). This enables quantum communication through existing fiber infrastructure without wavelength conversion. Entanglement can be distributed across metropolitan (10-100 km) or continental distances using existing dark fiber.

At 1550nm, standard single-mode fiber transmits photons with minimal loss. Erbium-doped fiber amplifiers (EDFAs) used in classical networks cannot amplify quantum signals (no-cloning theorem), but quantum repeaters based on entanglement swapping can extend range indefinitely.

Quantum Key Distribution (QKD)

QKD represents the first commercial application of photonic quantum technology. Companies like ID Quantique (Switzerland), Toshiba, and SK Telecom deploy QKD systems protecting financial transactions, government communications, and critical infrastructure. China has deployed the Beijing-Shanghai QKD backbone (2,000+ km) using trusted relay nodes.

Photonic quantum computers will eventually threaten current encryption (Shor's algorithm), but the same photonic technology provides the solution: QKD creates provably secure encryption keys based on quantum mechanics. Post-quantum cryptography and QKD together provide comprehensive security against both classical and quantum adversaries.

Quantum Repeaters

Long-distance quantum communication requires quantum repeaters to overcome fiber loss. Unlike classical repeaters that amplify and regenerate signals, quantum repeaters use entanglement swapping: creating entanglement between distant nodes by measuring and connecting intermediate entangled links.

Photonic approaches to repeaters include memory-based designs (storing photonic states in atomic or solid-state memory) and all-photonic designs (using measurement-based protocols without memory). PsiQuantum's architecture is compatible with both approaches, using the same silicon photonics manufacturing for repeater and computer components.

Molecular simulation represents the killer application for quantum computing, and photonic systems have particular advantages in this domain. Chemical reactions involve electronic transitions and molecular vibrations—processes that naturally map to photonic architectures. Vibronic spectra calculations, which describe how molecular vibrations couple to electronic transitions, are native to Gaussian Boson Sampling—Xanadu has demonstrated molecular simulation on their Borealis system with promising results.

Drug Discovery & Chemistry

Simulating molecular behavior remains intractable for classical computers beyond ~50 electrons. Quantum computers can simulate quantum mechanical systems directly, enabling accurate prediction of molecular properties, reaction rates, and drug-protein interactions. Pharmaceutical companies including Boehringer Ingelheim, Merck, and Roche have established quantum computing programs.

Near-term applications focus on hybrid classical-quantum algorithms: variational quantum eigensolvers (VQE) for ground state energy calculations, and quantum machine learning for molecular property prediction. Photonic systems excel at these hybrid workloads due to their connectivity and room-temperature operation enabling tight classical-quantum integration.

Optimization Problems

Combinatorial optimization pervades industry: supply chain logistics, financial portfolio optimization, scheduling, and network routing. These NP-hard problems scale exponentially for classical computers but may admit quantum speedups. Gaussian Boson Sampling provides advantages for graph-theoretic problems including maximum clique detection and densest subgraph identification.

PsiQuantum has partnerships with Mercedes-Benz for manufacturing optimization and with financial institutions for portfolio optimization. Xanadu's PennyLane framework enables developers to build optimization algorithms combining quantum and classical resources.

Machine Learning

Quantum machine learning (QML) represents a growing application area. Photonic systems can implement quantum kernel methods—computing kernel functions exponentially faster than classical methods for certain data structures. Xanadu's PennyLane framework integrates with PyTorch and TensorFlow, enabling hybrid quantum-classical neural networks.

Near-term QML focuses on: (1) quantum feature maps for classical data classification, (2) variational quantum circuits as trainable layers, (3) quantum generative models for sampling complex distributions. Photonic systems offer advantages for optical data (images, signals) where quantum and classical information processing can be seamlessly integrated.

Cryptography & Security

Quantum computers will eventually break RSA and ECC encryption using Shor's algorithm (requiring fault-tolerant systems with thousands of logical qubits). This drives urgent migration to post-quantum cryptography and quantum key distribution. The same photonic technology provides both threat (future code-breaking) and solution (QKD for secure keys).

Commercial QKD systems from ID Quantique, Toshiba, and others already protect financial transactions and government communications. Photonic quantum computers will eventually implement Shor's algorithm at scale, making this application timeline-dependent on when fault-tolerant systems achieve "cryptographically relevant" scale (estimated 2030s).

Path to Fault Tolerance

Fault-tolerant quantum computing requires logical qubits with error rates below threshold (~1% per operation). Physical qubits in any platform have higher error rates, so redundancy and error correction codes create protected logical qubits from many physical qubits. Both PsiQuantum's fusion-based approach and Xanadu's GKP states provide theoretical pathways.

PsiQuantum uses topological error correction combined with fusion gates. When fusion measurements create entanglement, failures are detected (heralded) rather than corrupting data. Combined with redundancy, this enables error correction without propagating failures. Xanadu's GKP states encode discrete information in continuous variables with built-in error correction—errors shift state positions, which can be measured and corrected.

Key Milestones

2022: Xanadu Borealis demonstrates quantum computational advantage with 216-qubit photonic processor. First photonic quantum computer on Amazon Braket cloud. Gaussian Boson Sampling solves in 36 microseconds what would take supercomputers 9,000 years.

2025: PsiQuantum Nature paper demonstrates fault-tolerance-capable components: 99.72% chip-to-chip fidelity. Xanadu achieves on-chip GKP state generation. Datacenter construction begins in Australia and Illinois. Xanadu announces $3B IPO.

2027: PsiQuantum commercial operation target. First datacenter-scale photonic quantum installations. Early logical qubit demonstrations expected. Integration of thousands of chips into unified systems.

2029+: Fault-tolerant photonic quantum computing. Million-qubit systems enabling practical quantum advantage for chemistry, optimization, and machine learning. Cryptographically relevant systems threatening current encryption standards.

Industry Investment

The photonic quantum computing sector has attracted massive investment. PsiQuantum leads with $700M+ in venture capital from BlackRock, Microsoft M12, Baillie Gifford, plus government partnerships worth $940M AUD from Australia. Xanadu's planned $3B IPO would represent the largest quantum company public market entry. Quandela and Photonic Inc. have raised over $100M each. Total sector investment exceeds $3B.

Government support is accelerating: Australia's National Quantum Strategy includes PsiQuantum facility funding. France's quantum initiative supports Quandela as a European champion. The EU's Quantum Flagship program has allocated €1B+ for quantum technologies including photonics. Japan and South Korea have national quantum programs with photonics components.

Manufacturing Scale-Up

PsiQuantum's datacenter-scale facilities represent the next phase: thousands of interconnected chips, millions of optical components, and the infrastructure for fault-tolerant computation. The Brisbane facility (Australia) and Illinois facility (US) will each house systems requiring dedicated buildings with specialized HVAC, power, and networking infrastructure.

Unlike superconducting systems where scaling requires exponentially more cryogenic capacity, photonic systems scale by adding modules at room temperature. Chip-to-chip interconnects enable modular growth—add more chips to the network for more qubits. This architectural advantage becomes critical at million-qubit scale.

Photon Loss: The Primary Challenge

The fundamental challenge for photonic quantum computing is photon loss. Unlike superconducting qubits which persist in circuits, photons can be absorbed, scattered, or fail to be detected at any point in the system. Each optical component introduces loss: single-photon sources operate at ~80% efficiency, waveguides attenuate signals at ~0.1 dB/cm, and detectors achieve ~90% efficiency in the best cases.

Cumulative loss limits circuit depth exponentially. A circuit with 1000 components, each at 99% transmission, has only 0.004% probability of the photon surviving. This is why photonic systems require loss-tolerant architectures. Fusion-based approaches (PsiQuantum) detect photon loss through heralding rather than allowing it to corrupt data. GKP states (Xanadu) encode information redundantly to tolerate photon loss within bounds.

Gate Fidelity & Probabilistic Operations

Two-qubit gates in photonic systems are inherently probabilistic—succeeding only 25-50% of attempts. This is fundamentally different from superconducting and trapped ion systems where gates succeed deterministically (though with finite fidelity). The KLM theorem proved this limitation is not fundamental: heralding identifies successful operations, and repeat attempts with fresh photons eventually succeed.

However, probabilistic gates limit overall clock speed. While each gate can be fast (photon propagation is effectively instantaneous), the need to repeat failed operations and the serial nature of some protocols means effective gate rates are MHz rather than GHz. Parallelism—running many operations simultaneously across spatial modes—partially compensates.

Clock Speed Comparison

Photonic quantum computers operate at MHz clock speeds (optical switching times, detector reset times, single-photon source rates) versus GHz for superconducting systems (nanosecond gate times). This 1000× difference in raw clock speed is often cited as a photonic disadvantage. However, the comparison is nuanced: photonic systems compensate through massive parallelism (thousands of simultaneous spatial modes) and connectivity (all-to-all rather than nearest-neighbor).

The relevant metric is time-to-solution, not clock speed. A slower clock with higher parallelism and better connectivity can solve problems faster than a faster clock with limited connectivity and gate fanout restrictions. For algorithms requiring long-range connectivity, photonic advantages in connectivity may outweigh clock speed differences.

Error Correction Overhead

Fault-tolerant quantum computing requires encoding logical qubits in many physical qubits. Surface codes (common for superconducting) require ~1000 physical qubits per logical qubit at current error rates. Photonic error correction codes (fusion networks, GKP states) have different overhead characteristics but similar orders of magnitude for resource requirements.

Neither photonic approach has demonstrated sub-threshold logical qubit error rates yet. PsiQuantum's components approach thresholds but haven't demonstrated full error-correcting circuits. Xanadu demonstrated on-chip GKP states but at fidelities below fault tolerance threshold. The next 2-3 years will determine whether photonics can cross this critical threshold.

Detector Cooling

While photonic systems operate at room temperature overall, single-photon detectors (superconducting nanowire detectors, SNSPDs) require 4 Kelvin cooling. This is far simpler than 15 millikelvin dilution refrigeration for superconducting qubits, but still requires cryogenic infrastructure. Research continues on room-temperature detector alternatives, but SNSPDs currently offer the best efficiency and timing performance.