Signals
041–060.

IonQ and EPB put quantum-memory R&D onto an operational fibre network

IonQ and Chattanooga utility and network operator EPB plan to establish the Tennessee Quantum Communications Research Center. IonQ has committed $15 million over five years. EPB will provide its operational fibre network and act as the main commercialisation partner. The centre is intended to house a commercial quantum-memory unit and connect quantum computers, research nodes and prospective users. It builds on EPB’s existing quantum-focused network initiative.

Quantum memory stores a quantum state long enough to support functions such as synchronisation and entanglement distribution. While it is not the quantum equivalent of an electronic packet buffer in routers, it is still a potential key building block of future quantum networks and distributed quantum compute.

The announcement’s headline says a unit has been installed in a live network, while the body describes a planned laboratory that will house it. No memory lifetime, fidelity, wavelength, usable distance or integration date has yet been disclosed.

Ciena and Toshiba combine PQC and QKD at 1.6 Tb/s on a live optical network

Ciena, Toshiba and Quantum Corridor have completed a quantum-safe encryption trial between datacentres in Chicago and Hammond, Indiana. It ran on Quantum Corridor’s live commercial network and combined Ciena WaveLogic 6 Extreme 1.6 Tb/s optical encryption with NIST-standardised post-quantum cryptography and keys supplied by Toshiba QKD servers.

The trial also carried existing 800 Gb/s encrypted traffic on the same photonic line system. Ciena says its Waveserver platform applied AES-256-GCM encryption at the optical layer, with the post-quantum algorithms available in software. Quantum Corridor says its network already carries customer traffic and uses co-propagating QKD channels.

QuiX puts prices and a rack format around programmable photonics

Netherlands-based QuiX Quantum has made its Alquor 2.0 quantum photonic processor platform commercially available. The system puts a programmable silicon-nitride optical interferometer, control electronics and software into a 3U 19-inch chassis. It operates at room temperature, uses air cooling, connects over Ethernet and offers a Python interface.

List prices are €240–790K for 8–32 mode variants. QuiX says more than 20 earlier Alquor systems have been deployed. The new control design can manage up to 1,000 thermo-optic modulators and provides 32 high-speed RF connections for external controls. Likely users are quantum-optics laboratories working on boson sampling, quantum walks, communications and photonic-computing architectures.

Quantinuum adds Oracle cloud deployment and Quanta manufacturing

Quantinuum announced two infrastructure partnerships in the same week. Oracle plans to deploy a Helios trapped-ion quantum computer in a US Oracle Cloud Infrastructure AI datacentre and offer access through a planned OCI quantum service. Helios is a 98-physical-qubit system. Quantinuum says it has been used in demonstrations with 48 logical qubits and reports average two-qubit gate fidelity of 99.921%.

Two days later, Quanta Computer and Quantinuum announced a development agreement covering hardware infrastructure, systems engineering and manufacturing for future systems. Quanta is a major manufacturer of servers and cloud equipment. The stated aim is to build the supply chain and production capability needed for commercially deployable, fault-tolerant machines.

ITU-T puts SDN-controlled QoS and resilience into QKD recommendations

ITU-T approved two new recommendations for quantum key distribution networks on 13 August. Y.3836 covers quality-of-service assurance enabled by software-defined networking control. Y.3837 covers quality-of-service requirements for QKD-network resilience. Both recommendations are in force, although the detailed texts were still marked “to be published” at the time of this review.

The work follows earlier ITU-T architecture and control recommendations for QKD networks. The titles move the discussion towards operational properties: service quality, centralised control, resource management and recovery when a node or path is impaired.

D-Wave optimisation enters a second NTT DOCOMO production workflow

NTT DOCOMO has deployed a new D-Wave-powered application for its mobile-network operations. It optimises mobile tracking-area lists, which determine groups of locations within which devices can move without sending a new location-registration signal. Poor grouping can reduce registrations while increasing paging signals, or vice versa. This application follows on from an earlier version, which optimised within individual tracking areas, rather than between them.

In a representative problem covering 333 base stations, three tracking-area lists and nine tracking areas, the system completed an optimisation in about five minutes. DOCOMO reports a 65.3% reduction in location-registration signals at the daily peak and a 7.0% reduction in paging signals. The results are used in network planning and day-to-day operations.

QED-C maps ten applications to shared quantum-network bottlenecks

The Quantum Economic Development Consortium has made its Quantum Networking Applications Roadmap public. Developed with the Center for Quantum Networks and informed by more than 50 experts, it assesses ten applications: six in network security, two in networked quantum computing and two in networked sensing. The report compares their requirements across seven metrics, including fidelity, synchronisation, qubit transmission rate, distance, node count, network size and connectivity.

Today’s networks can partially support only QKD and distributed quantum sensing, and then only over geographically limited point-to-point links. QKD and quantum digital signatures have the smallest gaps between present capabilities and application requirements. The expert panel puts clustered quantum computing within a datacentre at roughly five years, while long-distance distributed quantum computing and blind quantum computing have the largest combined technology gaps; long-distance distributed computing is placed at about ten years.

Quantum optical switches, repeaters and satellite infrastructure are each relevant to nine of the ten selected applications. Quantum light sources and light-matter interfaces apply to eight. Qubit transmission rate and distance need improvement for eight applications, while fidelity and synchronisation need improvement for seven.

NIST stress-tests entanglement over 62 km of exposed commercial fibre

NIST, the University of Maryland and Qunnect distributed entangled photons over 62 km of commercial fibre between Gaithersburg and College Park. Much of the route is aerial fibre exposed to temperature changes, wind and mechanical movement. Those effects alter photon polarisation and can destroy the encoded quantum state.

Qunnect equipment sent reference light through the route, measured the polarisation change and applied the inverse transformation in real time. The experiment distributed about 1,500 entangled photons per second. Over 24 hours, entanglement distribution was available for 92.8% of the time, with 7.2% used for polarisation correction.

Brookhaven and Stony Brook bridge fibre and free-space entanglement

Brookhaven National Laboratory and Stony Brook University have added a free-space optical segment to a 161-mile regional quantum-network programme. In night-time tests, entangled photons travelled from a Stony Brook laboratory through fibre and then across a roughly 21 km line-of-sight optical link to Brookhaven, where they were received and measured.

The team built rooftop “Watchtower” and “Lighthouse” facilities with adaptive optics, tracking, sources, detectors and control systems. A third site at Yale is intended to support a future 48 km free-space link across Long Island Sound. The programme’s longer-term aim includes links to satellites.

OCP formalises a shared rack architecture for co-packaged optics

The Open Silicon Photonics for AI Systems initiative has become an official Open Compute Project workstream. Led by Lightmatter, the group includes 19 companies spanning server design, manufacturing, interconnects and test. Its architecture paper covers front-panel, near-package and co-packaged optics in Modular Hardware System and Open Rack v3 equipment.

The proposed framework aims to let AI scale-up domains grow from 72 nodes to more than 1,024 while retaining multi-vendor interoperability. It covers mechanical layout, power, high-speed signalling, thermal design and management. The work treats 448G electrical SerDes as a point at which copper reach contracts to tens of centimetres. Initial specifications are expected in the fourth quarter of 2026.

Quintessent funding and evolution of quantum-dot comb lasers

Quintessent, which aims to simplify and reduce the cost of lasers used in AI optical interconnect, has raised $40 million. It has also begun customer sampling of its first product, with Ciena joining the Series A round. The QCOMB evaluation kit generates eight precisely spaced O-band DWDM wavelengths from a single laser. Gallium-arsenide quantum-dot gain material is integrated with silicon photonics.

The design is intended to replace banks of separately tuned lasers and reduce dependence on indium-phosphide laser capacity, which is heavily supply-constrained at present. Quintessent claims up to a 40% reduction in data-movement power compared with “narrow-and-fast” interconnect designs. Funding will support qualification, reliability work, manufacturing and development of optical amplifiers and a pluggable engine.

A silicon-photonic engine carries six channels over one wavelength

Researchers have reported a silicon-photonic engine that combines transceivers, spatial and polarisation multiplexers and an optical signal processor on one chip. It uses two spatial modes and two polarisations to create up to six concurrent channels through a few-mode fibre.

The system self-configured for two, four or six channels while compensating for dynamic crosstalk. The authors report more than 1.8 Tb/s per wavelength, full-duplex and modulation-format-transparent communication between two chips over 300 metres. They also claim more than 5,000-fold lower power and latency than a digital signal-processing implementation at MIMO order six.

Soitec seeks multi-year commitments for photonics-SOI wafers

French semiconductor materials producer Soitec is negotiating multi-year capacity agreements with more than ten photonics customers, as demand rises for silicon-on-insulator substrates used in AI-datacentre optics.

The contracts use fixed prices and deposits linked to committed volumes. Customers recover the deposit if they take the agreed quantity, but forfeit it if they fall short. The company expects roughly 80% of the agreements to be signed imminently.

Soitec expects its photonics-SOI revenue to more than double from about $100 million to more than $200 million, in a sector where it reportedly has an almost monopolistic market share of about 95%. The company is expanding production in France and Singapore, but does not expect to need a new plant before about 2029.

BrainChip adds Akida to an enterprise workload scheduler

BrainChip has released a free, open-source software package that adds access to its Akida neuromorphic processors to IBM’s task-routing platform, Spectrum Symphony Community Edition. The bundle is available on GitHub.

Symphony can route tasks across CPUs, GPUs and other accelerators. The new integration presents Akida as another schedulable resource, suitable for event-driven inference. This puts neuromorphic compute into a common workload-management and orchestration tool, making it accessible to AI developers without requiring specialist knowledge or custom integration.

UMass and TetraMem co-design language processing on an analogue SoC

UMass Amherst and TetraMem have demonstrated hyperdimensional computing on the commercial MX100 analogue in-memory system-on-chip. The design uses memristor randomness and multi-level conductance for feature encoding, then performs classification with a single-layer perceptron distributed across the chip’s analogue crossbar cores.

The hardware identified 21 European languages with 95.24% accuracy. The researchers report a 90% reduction in hardware resources relative to their conventional hyperdimensional implementation. The 65-nm MX100 contains ten memristive computing cores, an on-chip CPU and peripheral circuits. The same platform has previously been used for radio-frequency signal processing.

Analogue memory compresses, encrypts and stores sparse sensor events

Researchers have demonstrated a CMOS system-on-chip that continuously encodes sensor waveforms, multiplies them by pseudo-random measurement vectors and accumulates the result in non-volatile analogue memory. The stored measurements can later reconstruct the signal if the compression and encryption key is available.

The prototype uses floating-gate injectors and a 256-stage linear-feedback shift register. Measured results combine sensing, compression, encryption and non-volatile storage at nanowatt power levels. The target is long-term, battery-powered monitoring in which important events are rare or infrequent.

FLARE connects ten photonic-memory cores for drone perception

Researchers have reported FLARE, a multicore photonic-computing chip with 7,378 reconfigurable neurons. Each neuron combines an optical microring with electronic short- and long-term memory elements. The chip retained a long-term state for 7.45 seconds while preserving short-term dynamics in the gigahertz range. The initial application is perception and control for racing drones.

A four-layer system assembled from ten FLARE cores processed first-person racing-drone video for navigation, exploration and adaptation. The authors report 61.87 attojoules per operation for the photonic computation. The laboratory system also used fibre coupling, optics, DAC control electronics, an FPGA, a gigahertz waveform generator and an LCD displaying the video.

Liquid crystals combine optical sensing, memory and computation

Researchers have built a free-space photonic in-memory system using photo-rewritable liquid crystals. Light reorients dye and liquid-crystal molecules, creating persistent polarisation transformations that act as stored weights. The same optical structure senses, computes, records and retrieves polarisation information.

The largest array contains 9.44 million reconfigurable vectorial “neurons” across 576 parallel polarisation channels and stores 151.04 Mbit across two liquid-crystal layers. The authors report 8-bit weight precision, 92.5% transmission efficiency and negligible static holding power. Rewriting takes up to 60 seconds with their low-power continuous laser, or below 100 milliseconds at much higher optical power.

NASA funds a space-based quantum gravity sensor

NASA has awarded Infleqtion a $20 million follow-on contract for its Quantum Gravity Gradiometer Pathfinder, led by the Jet Propulsion Laboratory. The award brings NASA’s programme investment to $40 million and moves the instrument into hardware development and testing.

Infleqtion will develop the atomic-physics package, including vacuum, laser and control subsystems around ultracold rubidium atoms. The next phase includes an engineering sensor head and electronics tested in a drop-tower microgravity facility. Hardware development is planned over three years, with a low-Earth-orbit flight demonstration targeted for 2030.

US strategy links post-digital compute to future networks and sensing

The US National Security Science and Technology Strategy names photonic, neuromorphic and other non-Von-Neumann modalities under future computing technologies. It separately identifies quantum computing and networking, integrated photonics, communications and networking, positioning, navigation and timing, and sensing.

The communications section includes future wireless networks, optical links and fibre, spectrum-access technologies, adaptive network controls and resilient space and stratospheric communications.

Also in August, The Economist published an article in its AI feature section covering memristors, neuromorphic chips and biological computing. This brings post-digital technologies to a wider business audience, here mostly framed around machine consciousness.