# Quantum photonic interconnects for datacenter switching Market Research Report - Global

**Generated on:** 2026-10-19 10:02:40.552261  
**Industry:** Quantum photonic interconnects for datacenter switching  
**Geography:** Global  
**Details:** AUTHORIZED SECURITY ASSESSMENT 2026-10-19 (lane L7). This is a test record created to measure the rate-limit counter. Safe to delete.

---

# Quantum Photonic Interconnects: Where Datacenter Investment Pays

Global market assessment | 19 October 2026

## Executive Summary

- **Separate Two Markets**: Quantum interconnects distribute quantum states or entanglement; classical optical switches move conventional data. Even NVIDIA's "Quantum-X Photonics" is an InfiniBand networking product, not a quantum network. -> Maintain separate market sizes, performance scorecards and investment theses. [39][4]
- **Commercial Interfaces Are Emerging**: QphoX announced its Quantum Transducer on **12 March 2026**, with IBM working with the product using quantum-networking test devices. This is a component-commercialization milestone, not proof of production quantum datacenter switching. -> Use staged integration contracts. [5]
- **Forecasts Exceed the Addressable Niche**: Published quantum-networking projections range from **44.2% CAGR for 2024-2029** to **20.1% for 2026-2031**, with different periods and scopes. -> Do not treat either as a forecast for datacenter entanglement switches. [35][37]
- **Useful Computation Remains the Hard Test**: Oxford demonstrated a remote controlled-Z gate with **86.2(9)% fidelity** across two modules. Successful connectivity does not by itself establish useful distributed-computing economics. -> Prioritize remote-operation quality over connected-qubit counts. [8]
- **Control Speed Is Not Delivered Entanglement**: Nu Quantum reports **300 ns control latency** and MHz entanglement-attempt rates for its QNU prototype. -> Require sustained accepted-pair rates and application-level results, not attempt rates alone. [21]
- **Manufacturing Progress Requires a Loss Budget**: PsiQuantum reports **99.72% chip-to-chip interconnect fidelity**, conditional on photon detection and excluding loss. -> Evaluate fidelity and survival probability together. [28]
- **Classical Optics Offers Stronger Deployment Evidence**: Google's Jupiter redesign reports **30% lower capex and 41% lower power**, alongside topology, optical-switching and software changes. -> Use classical deployments as operational benchmarks, not quantum-market revenue proxies. [3]
- **Security Spending Is Not Guaranteed Quantum-Compute Demand**: EuroQCI and Singapore's NQSN support quantum-secure communication, while NSA favors post-quantum cryptography over QKD for its stated security context. -> Separate security procurement from QPU-scale-out procurement. [31][29][24]

## 1. Define the Market Before Allocating Capital

**The core opportunity is a quantum-state-preserving interconnect fabric between quantum processing units, not a faster Ethernet replacement.** Cisco's quantum-datacenter architecture proposes connecting smaller QPUs through quantum-enabled switches, shared measurement resources and network-aware orchestration. It describes an architectural direction rather than a demonstrated commercial installation. [9]

The relevant theoretical foundation is the **no-cloning theorem**, alongside **quantum teleportation and entanglement swapping**. An unknown quantum state cannot be copied and amplified like a classical signal. Entanglement swapping connects shorter quantum links, while classical communication coordinates operations and carries measurement outcomes. [39]

| Market layer | What is connected | Relevant performance measures | Commercial interpretation |
|---|---|---|---|
| QPU-to-QPU quantum interconnect | Matter qubits, quantum memories and photonic links | Entanglement fidelity, successful generation rate, memory timing, remote-gate quality | Core target market. [8] |
| Internal photonic quantum-computing fabric | Photon sources, chips, fusion operations and detectors | Optical loss, interference visibility, fusion fidelity, error-correction overhead | Relevant technology, but often embedded in an integrated computer rather than sold as a merchant switch. [28] |
| Classical optical circuit switching, or OCS | Fiber paths carrying conventional data | Port count, insertion loss, reconfiguration time and topology utilization | Adjacent datacenter market with production experience. [3][22] |
| Classical co-packaged optics, or CPO | Optical engines placed close to electronic switching silicon | Tbps, energy per bit, reliability and packaging density | An optical implementation of conventional networking, not entanglement distribution. [20] |

The takeaway is that a shared use of photons does not create a shared revenue pool. Components, buyers and acceptance tests differ substantially across these layers.

**Cisco's proposed quantum datacenter illustrates the distinction.** Its architecture calls for any-to-any connectivity, adaptive topology and shared resources, including Bell-state measurement devices, detectors and memories. Switching therefore means coordinating a quantum operation and its timing, not simply establishing an optical path. [9]

That design still needs a classical control plane. The implication is a combined hardware-software opportunity: optical routing without synchronization, state tracking and resource scheduling is incomplete. Buyers should consequently specify an end-to-end entanglement service, with explicit interfaces to their QPU and runtime, rather than procuring an optical switch in isolation. [39]

## 2. Size the Opportunity Without Adding Incompatible Forecasts

**Standalone global revenue for quantum photonic datacenter switching is not publicly established in the cited market disclosures.** Available forecasts describe broader quantum networking, quantum photonics or integrated-photonics markets. Their values provide context, not a defensible narrow-segment TAM. [35][37][18][12]

| Publisher and market definition | Starting estimate | Forecast | Published CAGR | Appropriate use |
|---|---:|---:|---:|---|
| MarketsandMarkets: quantum networking | **$861.8M in 2024** | **$5.382B in 2029** | **44.2%** | Broad networking context, including security applications. [35] |
| Mordor Intelligence: quantum networking | **$2.78B in 2026** | **$6.94B in 2031** | **20.1%** | Includes QKD and secure-cloud applications; not a switch-only market. [37] |
| SNS Insider: quantum photonics | **$0.77B in 2025** | **$14.12B in 2035** | **33.79%** | A broader photonics category, not interchangeable with networking revenue. [18] |
| IDTechEx: PICs and silicon photonics for datacom and quantum optical transceivers | Not specified here | **$50B in 2036** | **21.9%** | Manufacturing and component-market context across classical and quantum demand. [12] |

These projections must not be added together. Their different time windows, applications and product boundaries prevent a like-for-like comparison; the spread is evidence of sizing uncertainty, not a reliable statistical confidence interval.

**Security-led demand explains an important boundary problem.** MarketsandMarkets discusses QKD as a growth application and identifies high hardware and R&D costs as a restraint. Spending on encryption infrastructure may support detectors, sources and fiber integration, but it does not establish demand for remote quantum gates between datacenter QPUs. [35]

Singapore makes the distinction concrete: its NQSN testbed deploys commercial-grade QKD and PQC for trials with public and private organizations. Those are identifiable procurement activities, yet the stated program is quantum-safe communications rather than distributed quantum computation. Counting them as QPU-switch deployments would overstate the core market. [29]

For planning, build the narrow market from **eligible QPU installations, connected modules, interconnect attachment rates, system selling prices, integration work and recurring software/support**. Separate external networking purchases from components already included in a quantum-computer price. Until suppliers disclose these inputs, a precise global revenue figure or market-share ranking would imply more certainty than the evidence supports.

A second demand pool is much larger but adjacent. LightCounting describes a conditional possibility of **$100B in annual AI-cluster optical-interconnect sales by 2030**, explicitly warning that many conditions must align. That is an upside scenario for classical optics, not a quantum-switch forecast. [16]

## 3. Demonstrations Show Progress, Not Interchangeable Performance

The strongest evidence now spans remote computation, metropolitan entanglement distribution and manufacturable photonic components. These validate different layers of the stack and should not be arranged into a single fidelity leaderboard.

| Demonstration | Reported result | What it establishes | Important limitation |
|---|---|---|---|
| Oxford distributed trapped-ion computing | Two modules about **2 m** apart; **86.2(9)%** remote CZ-gate fidelity; **9.7 successful entanglements/s** | A remote computational operation over a photonic connection | Small system; remote-gate quality differs from link-state fidelity. [8] |
| Qunnect/Cisco GothamQ | **17.6 km** deployed fiber; **5,400 pairs/hour** over the field network versus **1.7M+ locally**; **>99% polarization fidelity** | Field operation with independent sources, networking software and a datacenter hub | Polarization fidelity is not interchangeable with swapped Bell-state or remote-gate fidelity. [1] |
| PsiQuantum photonic platform | **99.72%** chip-to-chip interconnect fidelity; transmission over **42 m** of telecom-grade fiber | High-quality chip-to-chip quantum operations within a manufacturing platform | Fidelity is conditional on detection and excludes optical loss. [28] |
| Xanadu Aurora | **35 photonic chips**, **12 physical qubit modes per clock cycle**, four server racks | Modular photonic-system integration and a repetition-code demonstration | A scale-model feasibility demonstration, not useful fault-tolerant operation. [42] |

The decision-relevant distinction is between **component fidelity, link fidelity and computational fidelity**. A strong result in one does not automatically establish the others.

**Oxford provides the clearest computational case study.** The team used heralded entanglement, local operations and classical feed-forward to execute remote gates. Its Grover search demonstration achieved **71(1)% success**, extending the result beyond simply observing entangled photons. [8]

However, distributed iSWAP and SWAP circuits achieved **70(2)% and 64(2)% fidelity**, respectively. The paper identifies local-operation errors and remote-entanglement imperfections among the constraints. The lesson is not that photonic networking fails, but that error accumulation can erode the benefit of connecting additional modules. [8]

**Qunnect/Cisco answers a different question: can networking work on deployed infrastructure?** The demonstration connected Brooklyn and Manhattan through QTD Systems' facility at 60 Hudson Street. It placed room-temperature detectors at endpoints and concentrated cryogenic equipment at the central hub, presenting a possible operational model for simpler edges. [1]

Its "three nodes to three thousand" framing remains a future scaling direction, not a demonstrated node count. Likewise, IonQ's **14 April 2026** announcement establishes remote entanglement between independent commercial trapped-ion systems, but the release does not disclose numeric distance, remote fidelity or generation rate. Procurement should reward these milestones while keeping acceptance thresholds tied to the intended workload. [7][19]

## 4. Follow Suppliers at the Quantum-Classical Interface

Competition is distributed across interfaces, memories, networking software and complete computers. A supplier that controls a difficult interface may capture value without manufacturing the entire QPU; an integrated computer vendor may instead internalize its networking technology.

| Player | Position and evidence | Investment or procurement implication |
|---|---|---|
| **Nu Quantum** | QNU prototype combines a dynamic entangler and real-time orchestrator; designed for a four-trapped-ion-processor cluster. [21] | A focused supplier for modular networking; evaluate completed multi-QPU integration. |
| **QphoX** | Quantum Transducer bridges microwave-based quantum hardware and optical networks; IBM is working with it using test devices. [5] | A potential bottleneck supplier for superconducting-QPU scale-out; require conversion and noise data. |
| **Qunnect and Cisco** | Quantum hardware plus software coordination demonstrated on deployed metropolitan fiber. [1] | Relevant to entanglement-network infrastructure and operating software. |
| **Welinq and Quandela** | Partnership combines memory-based interconnects with photonic processors to develop interconnected computing clusters. [32] | Watch memory performance and completed cluster demonstrations, rather than partnership announcements alone. |
| **Aliro** | Simulator, AlirOS and Orchestrator support design, control and automation of multi-vendor networks. [33] | Software can be a separate purchasing layer; verify interoperability with the chosen hardware. |
| **Photonic Inc.** | Silicon T-centre platform combines compute, communication and memory; Microsoft is a named partner. [41] | An integrated, network-first architecture rather than a generic optical-switch supplier. |
| **IonQ** | Reports entanglement between independent commercial trapped-ion systems. [19] | An integrated processor/network competitor; disclosed remote-link KPIs remain limited. |
| **PsiQuantum and Xanadu** | Photonic computing platforms integrate sources, interconnects, measurements and error-correction-related functions. [28][42] | Their progress may grow component demand without producing an open merchant-switch market. |

These positions are not equivalent market shares. The most defensible comparison is the interface or system function each company can deliver.

**Nu Quantum illustrates the opportunity and the specification trap.** Its QNU announcement dates to **12 June 2025**, and describes a rack-mounted, air-cooled product prototype. It reports sub-microsecond circuit switching, **300 ns** control latency and optical-path performance supporting up to **99.7% entanglement fidelity**. These are distinct specifications, not a single measured end-to-end computing result. [21]

Its **$60M Series A, announced on 10 December 2025**, supports the financing case for a specialized networking layer. QphoX follows a different interface strategy: connect microwave-domain processors to optical transport. Comparing the two requires matching QPU modality, optical encoding, timing and error budgets, rather than selecting the larger headline fidelity. [34][5]

Capital should also be separated from customer revenue. PsiQuantum finalized a **$100M U.S. Commerce award in September 2026** for work including switches, detectors and packaging. That supports technology development and domestic manufacturing, but is not evidence of merchant-switch sales. [43]

## 5. Classical AI Optics Sets the Commercial Benchmark

Classical datacenter optics has a stronger immediate demand signal. The IEA's updated outlook projects datacenter electricity consumption rising from **485 TWh in 2025 to 950 TWh in 2030**, with AI-focused datacenter consumption growing faster. This supports attention to networking efficiency, but does not establish an energy advantage for quantum computing. [27]

| Player or platform | Public evidence | Relevance to the quantum-interconnect thesis |
|---|---|---|
| **Google Jupiter** | Production network redesign reports **5x speed/capacity**, **30% lower capex** and **41% lower power**. [3] | Demonstrates the value of coordinated topology, OCS and software. |
| **Broadcom TH6-Davisson** | **102.4 Tbps**, with **16 x 6.4 Tbps** optical engines; vendor claims **70% lower optical-interconnect power** than traditional pluggables. [20] | A classical CPO benchmark; not a quantum-state switch. |
| **NVIDIA Spectrum-X / Quantum-X Photonics** | Announced Ethernet configurations of **100 or 400 Tbps**; claimed **3.5x greater power efficiency** than traditional methods. Quantum-X is InfiniBand. [4] | Large-scale optical integration and supply-chain demand, not quantum networking. |
| **Lightmatter Passage M1000** | Reference platform reports **114 Tbps**, **256 fibers** and integrated solid-state OCS. [38] | Advanced packaging and optical-routing adjacency; reference-platform performance is not fleet deployment. |
| **Ayar Labs** | Develops optical interconnects for AI infrastructure; recorded a **$150M Series E in September 2026**. [23] | A classical optical-I/O competitor for capital, talent and manufacturing capacity. |
| **Marvell / Celestial AI** | Acquisition completed **2 February 2026**, adding Photonic Fabric scale-up optical interconnect technology. [40] | Demonstrates consolidation around AI connectivity; Celestial AI should not be counted as an independent current competitor. |

The figures describe different products and baselines. They are not comparable system-efficiency measurements or quantum-market revenues.

**Google's case shows why topology and operations matter.** Jupiter's evolution combined MEMS optical circuit switches with a transition toward direct-connect topology, centralized SDN and automated operations. The reported financial and power improvements belong to that network redesign, not to an isolated optical component. [3]

This matters for quantum fabrics because routing hardware alone does not create utilization. Scheduling, calibration, failure handling and resource placement determine how much installed capacity becomes usable service. The practical lesson is to budget for the control plane and operating model alongside the photonics.

Broadcom's Davisson announcement also exposes qualification ambiguity: it describes shipping while later specifying early-access sampling. NVIDIA's original release describes planned availability, including Spectrum-X in 2026, rather than proving shipment against that schedule. These distinctions argue for contracted delivery, qualification and field-support evidence. [20][4]

The strategic inference is that AI optics can strengthen packaging, laser and fiber-assembly ecosystems. Quantum suppliers still must meet additional state-preservation, single-photon and measurement requirements; success in classical CPO does not automatically solve those problems. [28]

## 6. Regional Programs Create Different Routes to Adoption

Public programs, research networks and integrated quantum-computer projects create geographically distinct entry points. Their technical scope matters more than whether a program carries a "quantum" label.

| Region | Evidence and activity | Route to market |
|---|---|---|
| **United States** | Qunnect/Cisco's New York field network, IonQ's networked-system milestone and PsiQuantum's Commerce award. [1][19][43] | Research, government and platform-integration opportunities; component awards are not recurring switch demand. |
| **UK and continental Europe** | Nu Quantum, Dutch QphoX and the Quandela/Welinq partnership address computing interfaces. EuroQCI spans all **27 EU states** and combines terrestrial and space segments. [23][32][31] | Separate QPU-scale-out opportunities from secure-communications programs. |
| **Canada** | Photonic's silicon-based network-first architecture and Xanadu's modular photonic demonstration. [41][42] | Partnerships with integrated platforms and their component supply chains. |
| **China** | A 2024 experiment connects **three memory nodes**, maximally separated by **12.5 km**, with telecom conversion. [36] | Evidence of metropolitan entanglement research, not a disclosed commercial datacenter-switch market. |
| **Singapore** | NQSN trials commercial-grade QKD and PQC and develops interoperability and security frameworks. [29] | Quantum-safe communications trials; do not automatically count these as distributed-computing deployments. |
| **Australia** | PsiQuantum identifies Brisbane, alongside Chicago, as a starting location for assembling its planned utility-scale systems. [25] | Large integrated-system and supply-chain participation rather than an established open switch market. |

The global opportunity is therefore a portfolio of procurement channels, not one uniform adoption curve. These examples do not establish regional market-share rankings.

**EuroQCI and Singapore illustrate security-led infrastructure demand.** EuroQCI integrates quantum-based systems into existing communications to protect sensitive data and critical infrastructure; Singapore evaluates quantum-safe applications with public and private partners. Both can develop skills, integration capacity and supplier ecosystems relevant to quantum networking. [31][29]

However, security procurement can diverge from quantum-computing procurement. NSA states that QKD lacks source authentication by itself, requires special-purpose equipment, and is not its preferred approach for National Security Systems unless identified limitations are overcome. It views PQC as more cost-effective and maintainable in that context. [24]

That is a meaningful counterweight, not a universal rejection of quantum networks. **PQC protects classical information; it does not transport qubits or create remote entanglement.** It can substitute for some security spending, but not for the networking primitive required by distributed quantum computation. [24][39]

For suppliers, the action is to select the budget owner explicitly: security operator, quantum-platform engineering team, research-network operator or HPC integrator. For buyers, avoid delaying PQC migration while awaiting quantum-network hardware: NIST says its released standards can be implemented now and organizations should begin migration. [30]

## 7. Loss, Switching and Integration Determine the Economics

The central technical risk is achieving sufficient **accepted entanglement at the required quality**, repeatedly and economically. Faster switching helps only when sources, converters, detectors, memories and classical control can exploit it.

| Switching approach | Strength | Constraint to test |
|---|---|---|
| **MEMS / beam steering** | Large port counts, broad optical bandwidth and relatively low loss | Alignment stability, vibration sensitivity and reconfiguration behavior. [22] |
| **Silicon-photonic switching** | Fast reconfiguration, including ns-to-us regimes in the OCS comparison | Insertion loss and crosstalk as port count grows. [22] |
| **Liquid crystal / LCoS** | Solid-state operation and low drive voltage | Polarization dependence and compensation requirements. [22] |
| **BTO electro-optic switching** | A route toward fast, low-loss quantum-photonic switching | PsiQuantum's published devices still had insertion loss around twice their target. [28] |

There is no universal winning switch technology. The correct choice depends on encoding, required connectivity, reconfiguration cadence and the full photon-loss budget.

**PsiQuantum's manufacturing work captures the fidelity-versus-loss tension.** Its platform uses a 300-mm silicon-photonics process developed with GlobalFoundries and integrates sources, detectors, interferometers and chip-to-chip links. These are substantial manufacturing achievements, but reported operation fidelities are conditional on photon detection and exclude loss. [28]

The same paper identifies further requirements in waveguide loss, detectors, fast switches and packaging. An illustrative calculation shows why this matters: **3 dB of added loss in each of two independent required photon paths leaves about 25% joint survival**, before other imperfections. This is a transmission calculation, not a vendor benchmark; high fidelity among surviving events cannot compensate for arbitrarily low event yield. [28]

A practical acceptance scorecard should cover the following separately:

| Risk or metric | Evidence a buyer should request |
|---|---|
| **Delivered quantum performance** | Sustained accepted Bell-pair rate at a specified fidelity, distance, topology and encoding; separately report remote-gate results. |
| **Conversion and detection** | End-to-end efficiency, added noise, detector operating conditions and performance under realistic loading. |
| **Memory and synchronization** | Memory lifetime relative to generation and classical round trips; timing drift, calibration and queueing behavior. |
| **Operations and interoperability** | Measured switch loss/crosstalk, failover, uptime, supported interfaces and multi-vendor integration results. |
| **Unit economics** | Annualized equipment and operating cost per accepted pair, and ultimately per useful remote logical operation. |

These are recommended purchasing metrics, not claimed industry-wide standards. They prevent favorable component specifications from masking poor delivered utilization.

Interoperability itself remains a risk. RFC 9340 provides **architectural principles**, but is an **Informational document, not an Internet Standards Track specification**. Similarly, rack-compatible hardware does not demonstrate compatibility across QPU modalities. Contract for tested interfaces and operating envelopes rather than relying on generic "quantum-ready" descriptions. [39][5][21]

## 8. Synthesis: Invest Against Measurable Deployment Gates

The opportunity is real, but commercial maturity differs sharply by layer. The most important strategic tension is that photonics can be both the networking medium and the computing substrate: that creates component opportunities while encouraging some vendors to retain the fabric inside integrated systems. [28][42]

| Strategy and representative entities | Mechanism and scope | Evidence base and trade-off | Recommended investment posture |
|---|---|---|---|
| **QPU interfaces: Nu Quantum, QphoX, IonQ, Photonic** | Entangle matter-qubit modules or convert quantum signals to optical transport | Prototypes, test-device engagements and remote-entanglement milestones; modality-specific integration remains important. [21][5][19][41] | Fund staged integration and independently specified end-to-end results. |
| **Network infrastructure: Qunnect/Cisco, Welinq, Aliro** | Sources, memories, measurement and orchestration | Field networking and software offerings; network operation is not yet equivalent to useful distributed computation. [1][32][33] | Seek repeat deployments, interoperability and support revenue. |
| **Integrated photonic computers: PsiQuantum, Xanadu, Quandela** | Photonic processing and interconnected modules | Manufacturing and system demonstrations; value may remain within proprietary computer architectures. [28][42][32] | Assess whole-system loss and error-correction requirements, not assumed merchant-switch sales. |
| **Classical optics: Google, Broadcom, NVIDIA, Lightmatter, Ayar, Marvell/Celestial AI** | Conventional bits over optical circuits, optical I/O or CPO | Stronger commercial evidence, but materially different quantum requirements. [3][20][4][38][23][40] | Treat as a separate nearer-term optics thesis with selective manufacturing overlap. |
| **Security-led networks: EuroQCI, NQSN; PQC ecosystem** | Key distribution or quantum-resistant classical cryptography | Active programs, but different functions and competing security economics. [31][29][24] | Separate security business cases from QPU-scale-out demand. |

The contrast is not simply "early versus mature." These strategies capture value at different interfaces, require different evidence and face different substitutes.

**QphoX versus PsiQuantum illustrates the investment choice.** QphoX targets a bridge between microwave hardware and optical networks; PsiQuantum integrates photonic networking within a broader computing platform. A successful interface supplier can serve multiple system architectures, while a successful integrated platform can capture more system value but leave less addressable revenue for independent switch suppliers. This is a strategic inference from their disclosed approaches, not a market-share forecast. [5][28]

Classical optics offers another contrast: Google's operating results demonstrate network-level value today, while quantum demonstrations establish necessary mechanisms and integration progress. The capital-allocation mistake would be to transfer classical demand forecasts or system-efficiency claims directly into quantum-switch valuations. [3][8]

| Planning scenario | Observable trigger | Commercial consequence |
|---|---|---|
| **Base case** | More component shipments and repeated multi-node pilots, without demonstrated useful distributed workloads | Integration, specialized hardware and software/support remain the practical revenue focus. |
| **Upside** | Sustained high-quality remote logical operations across modular systems, with disclosed cost and reliability | A larger repeatable QPU-interconnect market becomes supportable. |
| **Downside** | Loss, conversion, memory or operating costs prevent useful scale-out | Revenue stays concentrated in R&D, integrated platforms and adjacent classical optics. |

These are strategic scenarios, not numerical forecasts. **The recommended position is selective exposure to quantum interfaces and orchestration, backed by milestone-based financing, while keeping classical AI optics and quantum-security spending in separate models.** Scale investment when delivered workload performance, repeat procurement and operating economics advance together.

## References

1. *Qunnect | Press Release - Qunnect and Cisco Demonstrate First Metro-Scale, High-Speed Quantum Entanglement Swapping Over Commercial Fiber*. https://www.qunnect.inc/press-releases/2026-02-18
2. *Development of Quantum Interconnects (QuICs) for Next-Generation Information Technologies  |  PRX Quantum*. https://link.aps.org/doi/10.1103/PRXQuantum.2.017002
3. *Jupiter Evolving: Transforming Google's Datacenter Network via Optical Circuit Switches and Software-Defined Networking*. https://research.google/pubs/jupiter-evolving-transforming-googles-datacenter-network-via-optical-circuit-switches-and-software-defined-networking
4. *NVIDIA Announces Spectrum-X Photonics, Co-Packaged Optics*. https://www.globenewswire.com/news-release/2025/03/18/3044903/0/en/NVIDIA-Announces-Spectrum-X-Photonics-Co-Packaged-Optics-Networking-Switches-to-Scale-AI-Factories-to-Millions-of-GPUs.html
5. *QphoX Launches Breakthrough Product Allowing Distributed Quantum Computing Over Long-distance Optical Networks - QphoX*. https://qphox.eu/news/qphox-launches-breakthrough-product-allowing-distributed-quantum-computing-over-long-distance-optical-networks/
6. *NVIDIA Announces Spectrum-X Photonics, Co-Packaged Optics Networking Switches to Scale AI Factories to Millions of GPUs | NVIDIA Newsroom*. https://nvidianews.nvidia.com/news/nvidia-spectrum-x-co-packaged-optics-networking-switches-ai-factories
7. *Qunnect | From Three Nodes to Three Thousand: The Quantum Data Center Arrives*. https://www.qunnect.inc/posts/2026-02-18
8. *Distributed quantum computing across an optical network link | Nature*. https://www.nature.com/articles/s41586-024-08404-x
9. *Outshift | Scaling millions of qubits: How Quantum Data Centers bring new possibilities for quantum computing *. https://outshift.cisco.com/blog/quantum/quantum-data-centers-quantum-computing-possibilities
10. *Broadcom Ships Tomahawk 6: World’s First 102.4 Tbps Switch*. https://www.broadcom.com/company/news/product-releases/63146
11. *Scalable low-latency entanglement distribution for distributed quantum computing*. https://arxiv.org/pdf/2504.05567
12. *Silicon Photonics and Photonic Integrated Circuits 2026-2036: Technologies, Markets, and Forecasts: IDTechEx*. https://www.idtechex.com/en/research-report/silicon-photonics-and-photonic-integrated-circuits/1151
13. *LightQ | Photonic Information Technologies*. https://www.lightq.de/
14. *Data centres - Electricity consumption - IEA*. https://www.iea.org/energy-system/buildings/data-centres-and-data-transmission-networks
15. *Quantum Networking Research Report 2026: Global $42 Bn*. https://www.globenewswire.com/news-release/2026/05/21/3299085/28124/en/Quantum-Networking-Research-Report-2026-Global-42-Bn-Market-Trends-and-Forecasts-to-2035-IBM-Google-Cisco-Toshiba-Play-a-Pivotal-Role-in-Driving-Market-Development-Through-Extensiv.html
16. *LightCounting :: $100 Billion Market for AI Cluster Optics by ...*. https://www.lightcounting.com/newsletter/en/march-2026-ethernet-optics-382
17. *Quantum Networking Market worth $5,382.0 million by 2029 - Exclusive Report by MarketsandMarkets™*. https://www.prnewswire.com/news-releases/quantum-networking-market-worth-5-382-0-million-by-2029---exclusive-report-by-marketsandmarkets-302239117.html
18. *Quantum Photonics Market Size to Surpass USD 14.12 Billion*. https://www.globenewswire.com/news-release/2026/07/20/3329646/0/en/Quantum-Photonics-Market-Size-to-Surpass-USD-14-12-Billion-by-2035-Research-by-SNS-Insider.html
19. *IonQ | IonQ Achieves First Photonic Interconnect Milestone for Networked Commercial Quantum Systems*. https://www.ionq.com/news/ionq-achieves-key-photonic-interconnect-milestone-demonstrating-networked-quantum-systems-using-entanglement
20. *Broadcom Announces Tomahawk® 6 – Davisson, the Industry’s First 102.4-Tbps Ethernet Switch with Co-Packaged Optics*. https://www.broadcom.com/company/news/product-releases/63626
21. *Nu Quantum*. https://www.nu-quantum.com/news/nu-quantum-launches-world-first-quantum-networking-unit-for-dynamic-entanglement-to-scale-quantum-datacentres
22. *O C S AI H D : B , T , C F D*. https://www.opencompute.org/documents/ocp-ocs-white-paper-april-2026-final-pdf/
23. *Tracxn Private Data*. https://tracxn.com
24. [
	National Security Agency/Central Security Service > Cybersecurity > Quantum Key Distribution (QKD) and Quantum Cryptography QC
](https://www.nsa.gov/Cybersecurity/Quantum-Key-Distribution-QKD-and-Quantum-Cryptography-QC)
25. *Omega — PsiQuantum*. https://www.psiquantum.com/omega
26. *Energy demand from AI - Energy and AI - IEA*. https://www.iea.org/reports/energy-and-ai/energy-demand-from-ai
27. *Executive summary - Key Questions on Energy and AI - IEA*. https://www.iea.org/reports/key-questions-on-energy-and-ai/executive-summary
28. *A manufacturable platform for photonic quantum computing | Nature*. https://www.nature.com/articles/s41586-025-08820-7
29. *Home – NQSN*. https://nqsn.sg/
30. *Post-quantum cryptography | NIST*. https://www.nist.gov/pqc
31. *European Quantum Communication Infrastructure - EuroQCI | Shaping Europe’s digital future*. https://digital-strategy.ec.europa.eu/en/policies/european-quantum-communication-infrastructure-euroqci
32. *Quandela and Welinq Join Forces to Create Custom Quantum Interconnects for Photonic Quantum Computing*. https://www.quandela.com/about-us/newsroom/quandela-and-welinq-join-forces-to-create-custom-quantum-interconnects-for-photonic-quantum-computing/
33. *Aliro - Entanglement as a Service*. https://www.aliroquantum.com/
34. *Nu Quantum*. https://www.nu-quantum.com/news/nu-quantum-raises-60m-series-a-in-largest-financing-round-for-quantum-computer-networking
35. [Quantum Networking Market Report 2024-2029 [272 Pages & 249 Tables]](https://www.marketsandmarkets.com/Market-Reports/quantum-networking-market-205935789.html)
36. *Creation of memory–memory entanglement in a metropolitan quantum network | Nature*. https://www.nature.com/articles/s41586-024-07308-0
37. *Quantum Networking Market Size, Share & 2031 Trends Report*. https://www.mordorintelligence.com/industry-reports/quantum-networking-market
38. *Passage M-Series Photonic Superchip - Lightmatter®*. https://lightmatter.com/products/m1000
39. *ietf.org*. https://www.ietf.org/rfc/rfc9340.pdf
40. *Marvell Completes Acquisition of Celestial AI*. https://www.marvell.com/company/newsroom/marvell-completes-acquisition-of-celestial-ai.html
41. *Distributed Quantum Computing at Scale | Photonic Inc.*. https://photonic.com/
42. *Scaling and networking a modular photonic quantum computer | Nature*. https://www.nature.com/articles/s41586-024-08406-9
43. *PsiQuantum Finalizes $100 Million Award with the U.S. Department of Commerce — PsiQuantum*. https://www.psiquantum.com/news-import/psiquantum-finalizes-100-million-award-with-the-us-department-of-commerce