The convergence of Quantum with AI
Quantum information encoded in light. Operates at room temperature and integrates naturally with fibre networks, but deterministic two-qubit gates and photon loss remain the central engineering problems.
94 stories
The Quantum Internet Alliance has received €47.5 million from the European Commission. The funding is intended to support construction of a full-stack quantum network prototype.
This could make a metropolitan-scale quantum network testbed available to early application developers within two years.
A preprint posted to arXiv introduces QuLoC, a compression pipeline for large language models that combines SVD-based low-rank approximation with outputs from photonic quantum circuits. The quantum outputs are used as gating signals intended to preserve downstream task performance after parameter reduction. The abstract describes the method but does not report experimental benchmarks or hardware demonstrations.
If QuLoC can be simulated on classical hardware or run on small photonic processors, it could produce a benchmark showing improved downstream accuracy over standard low-rank SVD on a small language model within two years.
A University of Vienna and VCQ team has operated a programmable quantum photonic processor aboard a nanosatellite in orbit. The system demonstrated two-photon quantum interference and programmable matrix operations. This marks a shift from spaceborne quantum communication to active on-board quantum processing at the edge.
If entangled photon sources can be integrated, this demonstration could allow nanosatellite constellations to perform on-orbit entanglement swapping and routing, reducing dependence on ground stations for secure key distribution.
An arXiv preprint describes a Bayesian likelihood-free inference method that uses deep learning to estimate parameters from photodetector click patterns with non-classical statistics. The work is aimed at enabling quantum sensors to operate in real time at the quantum limit. It addresses photodetection, where fast interpretation of such patterns is a bottleneck.
This could allow quantum sensors to process non-classical photodetection data in real time on conventional computing hardware, making quantum-enhanced measurements feasible for continuous field operation.
Researchers posted a paper describing a hybrid photonic architecture built on thin-film lithium tantalate, intended to serve as nodes in long-distance quantum networks. The design pairs the lithium tantalate platform with atomic frequency comb quantum memories. The authors present it as a scalable alternative to diamond-based integrated photonics, whose integration has remained difficult.
Within two years, this architecture could enable prototype quantum network nodes where photonic routing and quantum memory sit on the same chip, reducing the interface losses that currently limit entanglement distribution.
A preprint on arXiv analyzes teleportation-based Gottesman-Kitaev-Preskill (GKP) error correction when the ancilla Bell pair is not decoherence-free, focusing on pure-dephasing processes that commonly affect bosonic systems. The authors examine how this realistic noise source degrades protocol performance, moving beyond the usual idealization of perfect ancilla resources.
The analysis could give experimental groups quantitative dephasing thresholds for GKP ancilla states, letting them set realistic coherence targets for near-term bosonic-qubit demonstrations.
Researchers have proposed a method for generating photonic graph states that minimises the number of quantum emitters required. The work addresses the challenge of deterministic preparation of these states, which are used in measurement- and fusion-based quantum computing, quantum networks, and sensing. It also engages with recent efforts to find efficient preparation schemes, including heuristic optimisation approaches.
This could allow photonic quantum computing groups to build small graph-state sources with fewer quantum emitters, reducing the hardware overhead for near-term fusion-based prototypes.
An arXiv preprint reports a quantum advantage in the sample complexity of learning single-mode bosonic channels. The authors frame the result against earlier exponential quantum learning speedups that required quantum resources, such as system dimension or entanglement, to grow with the problem. The new work concerns whether such scaling is necessary for continuous-variable channel estimation.
The result becomes a benchmark for classical simulation and adaptive measurement strategies, prompting efforts to match or refute the claimed speedup.
Photonic Inc. and Microsoft are working together to add Photonic's distributed-computing models and quantum low-density parity-check error-correction routines to Microsoft's Quantum Resource Estimator. The aim is to let researchers and developers estimate computational overheads more accurately for algorithms running on multi-chip, distributed quantum systems. The work addresses limitations of existing resource estimators that were designed primarily for monolithic architectures.
If the integration lands as described, algorithm design teams could start using Microsoft's Quantum Resource Estimator within the next two years to compare distributed and monolithic hardware trade-offs for early fault-tolerant quantum algorithms, making modular architectures a default scenario in planning tools rather than a niche add-on.
Researchers report an ultra-low-power integrated photonic reservoir computer aimed at quantum machine learning. Their approach is designed to introduce nonlinearity and short-term memory into photonic computation without active tuning or additional nonlinear elements. The authors position the platform as versatile for machine learning tasks.
Within two years, integrated photonic reservoir chips of this kind could become a candidate backend for low-power edge inference on time-series or signal-classification tasks where milliwatt-level operation is decisive.
The paper presents a single-atom-based photonic interconnect intended to link modular quantum processors. It addresses entanglement distribution across optical channels and positions the scheme relative to loss-resilient protocols built on linear-optics type-II fusion gates.
If the single-atom memory can absorb and re-emit photons on demand, small quantum modules could be linked without nanosecond-scale synchronisation within two years.
A preprint proposes a quantum machine learning framework in which multiple quantum states are loaded in parallel so the model can learn from relationships between states, not just individual instances. The authors describe an adaptive relational learning method that captures pairwise and higher-order structure in multi-instance quantum data, targeting photonic processors. The abstract does not report experimental results or hardware demonstrations.
If the proposed relational encodings can be implemented on near-term photonic hardware, this could make photonic QML models practical for graph- or set-structured quantum datasets where pairwise and higher-order correlations are the signal.
A new arXiv preprint presents a security proof claiming unconditional security for discrete-modulated continuous-variable quantum key distribution. The work is framed as closing a gap between the protocol's practical advantages in telecom infrastructure and the lack of rigorous security guarantees.
If the proof is validated, discrete-modulated CV-QKD systems could be certified and deployed in metropolitan networks without relying on Gaussian-modulation assumptions, because the theoretical barrier would be removed.
A new arXiv preprint proposes a fault-tolerant computing scheme in which physical qubits are spread across multiple processors connected by photonic interconnects. The authors argue this distributed design can make error correction more efficient and improve hardware scalability. The abstract does not report experimental results or specify the underlying qubit technology.
Within two years, this framework could become a design reference for modular quantum hardware programs, such as IBM's multi-chip superconducting processors or QuEra's networked neutral-atom arrays, seeking to demonstrate a logical qubit spanning separate modules.
An arXiv preprint reports the experimental demonstration of a logical Bell-state measurement that exceeds the linear-optical limit. The paper frames the result within fault-tolerant quantum computing, where Bell-state measurements are building blocks for measurement-based and fusion-based quantum computation and for quantum networks.
If the logical Bell-state measurement can be integrated with existing photonic encodings, it could allow fusion-based photonic quantum processors to replace standard linear-optical fusion operations with higher-success logical variants, reducing the overhead required for fault-tolerant operation within the next two years.
Xanadu and AMD have released Backline, an open-source extension for PennyLane designed to link quantum processors to classical compute resources including CPUs, GPUs, FPGAs, and SmartNICs. The framework provides Python-native, microsecond-scale communication aimed at removing the data bottleneck between classical and quantum systems for workloads such as quantum error correction.
Backline could make real-time quantum error correction experiments practical on near-term quantum processors by supplying microsecond-latency feedback between quantum hardware and classical decoders.
Researchers have shown that BB84 encryption achieves a stronger security property than previously proven. Earlier unclonable encryption schemes only guaranteed 'search' security, where an attacker could obtain some valid key. The new result establishes 'unclonable indistinguishability', meaning no adversary can tell encrypted messages apart.
This proof could push standards bodies to adopt unclonable indistinguishability as a required security definition for commercial QKD products within two years.
Xanadu and AMD have introduced Backline, an open-platform interface intended to move data between quantum processors and classical CPUs, GPUs, and FPGAs with latency measured in microseconds. The design is presented as a way to support tighter coupling of quantum and classical compute for hybrid workloads.
Within two years, Backline could allow Xanadu's photonic processors to use AMD FPGAs for low-latency feedforward operations, such as conditional state preparation or correction based on measurement outcomes, without a host-PC round trip.
PsiQuantum has finalized a $100 million award with the U.S. Department of Commerce. The funding is intended to strengthen domestic quantum computing and semiconductor security.
This could lead to a U.S.-based pilot production line for PsiQuantum's photonic quantum components within two years, reducing its dependence on overseas semiconductor fabrication.
PsiQuantum finalized a $100 million CHIPS Act award with the U.S. Department of Commerce to accelerate domestic semiconductor process development for fault-tolerant quantum computing. The funding is directed at 300mm wafer production of barium titanate and optimization of single-photon sources.
PsiQuantum could bring barium titanate photonic component fabrication onto 300mm wafers at a domestic facility within two years, giving it an onshore supply of single-photon sources for iterative device testing.