Quantum Networks

Established A quantum network distributes quantum states — above all, entanglement — between physically separate nodes, and this is a different and far harder task than quantum key distribution (QKD), which merely uses quantum signals to negotiate classical secret keys between two points. The two get sold under one name, and the conflation is the single most common error in coverage of the field. QKD hardware has been commercially available for two decades and China operates a 4,600-km trusted-relay QKD network; genuine entanglement distribution between independent quantum devices has, as of 2026, been demonstrated inside a handful of metropolitan areas at rates between one success per hour and a few per second. This brief keeps the two ledgers separate throughout.

Established The no-cloning theorem forbids amplifying an unknown quantum state, so the classical repeater — the amplifier that made planetary telecommunications possible — is illegal physics for qubits. The best deployed telecom fibre attenuates light by roughly 0.17–0.2 dB per kilometre, so photon survival falls exponentially with distance: after 400 km, fewer than one photon in a hundred million arrives. Pirandola, Laurenza, Ottaviani and Banchi proved in 2017 (the PLOB bound) that without intermediate stations the achievable rate of any point-to-point quantum link scales linearly with channel transmittance — a hard ceiling, not an engineering shortfall. The escape is the quantum repeater, proposed by Briegel and colleagues in 1998 and given its canonical atomic-ensemble form by Duan, Lukin, Cirac and Zoller (DLCZ) in 2001: divide the line into segments, entangle each segment heralded by photon detection, hold the entanglement in quantum memories, and join segments by entanglement swapping.

Established Wehner, Elkouss and Hanson’s 2018 roadmap in Science defines the accepted ladder: trusted-repeater QKD sits at stage zero precisely because it is not quantum end to end, and the true stages climb from prepare-and-measure links through entanglement distribution, memory networks, and eventually networked fault-tolerant processors. By that ladder, deployed commerce sits at stage zero, the leading laboratories sit at stage two to three inside single cities, and everything above is design.

Frontier As of September 2026 every component of a quantum repeater exists somewhere, and the assembled machine exists nowhere. Memories have stored quantum states for an hour; entanglement has crossed 420 km of fibre loss; multiplexing has raised rates fifteenfold; a memory has beaten direct transmission once, inside a single laboratory. No experiment has yet combined them into a repeater that outperforms a plain glass fibre in the field. That scarcity, twenty-eight years after the proposal, is the central finding of this brief.

Established Point-to-point QKD over real infrastructure is a solved engineering problem at metropolitan-to-regional scale. The current field benchmark is Toshiba Europe’s 2025 Nature demonstration (Pittaluga and colleagues) of coherent twin-field QKD over 254 km of commercially operated Deutsche Telekom fibre between Frankfurt and Kehl (vendor-led research): secret key at roughly 110 bits per second across the full span, using room-temperature semiconductor detectors in standard telecom racks rather than cryogenically cooled superconducting ones — the first time coherent, phase-stabilised quantum communication ran on national-scale commercial infrastructure without a physics-lab environment. The laboratory distance record for the same protocol family is 1,002 km of spooled ultralow-loss fibre (2023), at 0.0034 bits per second — a working link, and a demonstration that rate collapses with distance even for the best repeaterless protocols.

Established China’s national network, the largest quantum-communication deployment on Earth, is a chain of classical trust, not a quantum network. The 2,000-km Beijing–Shanghai backbone (operational 2017) and the integrated space-to-ground system described in Nature in 2021 — roughly 4,600 km of reach, more than 150 institutional users, two satellite ground links — relay keys through dozens of intermediate stations, each of which decrypts and re-encrypts. Every relay is a point where the key exists in the clear; the security model is guards and vetting, not physics. The Micius satellite’s 2017 result stands apart as real entanglement: photon pairs distributed to ground stations 1,203 km apart at a detected rate near one pair per second, violating a Bell inequality — still the terrestrial-distance record for entanglement of any kind, achieved by going over the atmosphere rather than through glass.

Established Western security agencies have formally declined to endorse QKD, and their objections are technical, not rhetorical. The US National Security Agency states that it does not recommend QKD for securing National Security Systems, citing five limitations: QKD is only a partial solution (it still requires authenticated classical channels), demands special-purpose hardware, raises infrastructure cost and insider-threat exposure through trusted relays, is hard to certify because security rests on physical implementation rather than testable mathematics, and increases denial-of-service risk. The UK National Cyber Security Centre’s Quantum security technologies white paper (2020) states it does not endorse QKD for any government or military applications and points instead to post-quantum cryptography (PQC). A 2024 joint position paper by the German BSI, French ANSSI, Dutch and Swedish national agencies reached the same conclusion: PQC first, QKD immature. NIST finalised its first PQC standards (ML-KEM and companions) in August 2024, and they are already deployed in mainstream browsers — software that neutralises the store-now-decrypt-later threat QKD is marketed against, at approximately zero marginal hardware cost. The key-distribution market that funded the field’s first two decades is contested by its own governments.

Frontier Entanglement distribution without memories is now an operations story rather than a physics story. Qunnect’s GothamQ link (PRX Quantum, 2024) (vendor’s own figures) distributed polarisation-entangled photon pairs over 34 km of deployed New York City fibre for 15 continuous days at 99.84 percent uptime, generating roughly 500,000 pairs per second at the source and maintaining about 99 percent fidelity at rates up to 20,000 pairs per second, with fully automated polarisation compensation. Photon pairs alone cannot store or route anything, but this is the rate regime — four to six orders of magnitude above every memory experiment — that memory-based networks must eventually reach.

Frontier Memory-to-memory entanglement crossed from laboratory corridors to city fibre in 2024, in three independent demonstrations on two continents. In Hefei, the USTC group of Pan and Bao built a three-node network with atomic-ensemble memories and telecom conversion, nodes up to 12.5 km apart, with entanglement generated between any two nodes through a central photonic server (Nature, May 2024). In Boston, the Harvard group entangled two silicon-vacancy diamond memories through a 35-km deployed fibre loop crossing four municipalities: Bell-state fidelities of 0.86 and 0.74 for the electron spins, nuclear-spin memory entanglement at 0.77 with error detection, heralding success around 1 Hz, bidirectional frequency conversion between 737 nm and the 1,350 nm telecom O-band at 5.4 percent efficiency, both nodes below 200 mK in dilution refrigerators (Nature, 2024). In the Netherlands, the QuTech consortium (Stolk and colleagues, published 2025) heralded entanglement between nitrogen-vacancy processors in Delft and The Hague across roughly 25 km of deployed fibre — the first fully heralded link between independently operated solid-state nodes in different cities.

Frontier The 2025–2026 results attack the two known weaknesses: rate and certification. A Hefei trapped-ion node (Cui and colleagues, Nature Communications, December 2025) used hybrid multiplexing across a calcium-ion chain to reach 4.28 ion-photon entanglement events per second at 89.8 ± 1.1 percent fidelity over 12 km, a 15.6-fold multiplexing gain, with 866-to-1,558 nm conversion at 12 percent efficiency. The XingHan 2.0 system (Zhu and colleagues, Nature Photonics, May 2026) entangled multiplexed rare-earth solid-state memories 14.5 km apart over China Unicom fibre at 78.6 ± 2.0 percent fidelity and certified the result with a CHSH Bell violation, S = 2.22, 3.7 standard deviations above the classical bound — the first metropolitan memory-to-memory link whose entanglement is certified device-independently rather than inferred from tomography.

Frontier The distance record — 420 km, reported by USTC in Physical Review Letters in August 2026 — is a channel-length benchmark, not a network. Two rubidium-ensemble memories were entangled through 420 km of fibre totalling 78.7 dB of loss, at 1.09 successes per million attempts and a concurrence of 0.046 ± 0.022, with 780-to-1,522 nm telecom conversion and a dual-band phase-stabilisation scheme. The caveats matter as much as the number: both memories sat in the same laboratory, only 20.2 km of the fibre was deployed underground (the remaining 400 km was spooled ultralow-loss fibre), and the memories were read out 750 nanoseconds after writing — before the heralding signal could possibly return — so the entanglement is verified after the fact rather than stored and used. What the experiment does prove is significant: beyond roughly 230 km its rate scales with the square root of transmittance rather than linearly, the first memory-based system to beat repeaterless scaling at such losses. It is a demonstration that the repeater arithmetic works, not that a repeater exists.

Established Across every platform, the measured record over long or deployed fibre spans rates from one success per 72 minutes to a few per second, at fidelities between 53 and 90 percent. The 2025 review by Zhu, Liu, Zhou and Li collects the numbers: rubidium ensembles at 22 km, 73.2 ± 3.8 percent fidelity near 0.007 Hz (2020); single atoms through 33 km of coiled fibre at 62.2 percent (2022); NV centres over 10 km at 53.4 percent (2024); trapped ions over 230 m at 88 percent (2023); silicon-vacancy centres over the 35-km urban loop at 69 ± 7 percent and 0.00023 successes per second (2024). Entangled pairs and classical bits are not commensurable, but the internal comparison is stark: memory networks run four to six orders of magnitude below the photonic-pair rates already demonstrated on the same kind of fibre, and repeater architectures need thousands of pairs per second to be worth building.

The open questions sort into what must be discovered and what must merely be built.

Frontier Can any platform pass repeater break-even in the field? The only demonstration anywhere of a quantum memory increasing a communication rate beyond the direct-transmission bound is Bhaskar and colleagues (Harvard, Nature, 2020) — a single silicon-vacancy memory acting as a two-way interface inside one laboratory. No experiment has repeated this over deployed fibre between separate sites. This is the field’s acknowledged gate, discussed further under Required experiments.

Frontier Which memory platform wins is genuinely open, because each owns one corner of an incompatible triangle. Atomic ensembles herald efficiently (76 ± 5 percent readout) but store for subseconds; trapped ions and NV nuclear spins store for minutes to an hour but entangle at hertz or below; rare-earth crystals multiplex hundreds of modes and one europium crystal has held nuclear-spin coherence for six hours (Zhong and colleagues, Nature, 2015) but retrieval efficiencies remain low; Qunnect’s warm rubidium vapour needs no cryogenics and speaks 1,324 nm natively, but stores for only 3 microseconds. No platform yet combines second-scale storage, better-than-half efficiency, and telecom compatibility — the combination a repeater chain actually requires.

Frontier Can superconducting quantum computers join a network at all? Their qubits live at gigahertz frequencies; fibre carries 200-terahertz photons. Microwave-to-optical transduction is the required bridge, and reported end-to-end efficiencies remain at the percent level with added noise near one photon (Han and colleagues’ Optica review, 2021, and successors); no device simultaneously achieves high efficiency, low noise, and useful bandwidth. Until transduction improves by roughly two orders of magnitude, or optically networked platforms (ions, neutral atoms, diamond) win the processor race, “networked quantum data centres” remain a slide, not a plan.

Frontier How far can multiplexing carry the rate problem? The 15.6-fold ion-chain gain and the temporal multiplexing in XingHan 2.0 suggest rates scale with mode count roughly as theory predicts; whether thousand-mode memories preserve fidelity is untested.

Speculative Entanglement routing — scheduling, buffering, and swapping across a mesh of lossy, expiring links — has a literature and no experiment. IETF RFC 9340 (2023) lays out architectural principles, and simulators exist at several institutes, but no three-city network has ever routed entanglement between endpoints that lack a direct link. The protocols may need revision the first time they meet real hardware jitter.

Speculative Does sensing pay before computing does? Entangled clock networks and long-baseline interferometry are the applications with plausible early customers (metrology institutes, radio astronomy), but no networked-sensing demonstration has yet beaten the best classical alternative at any task with an external user.

Established The binding bottleneck is the quantum memory triangle: storage time, efficiency, and telecom compatibility, of which every existing device delivers at most two. A repeater segment of 100 km imposes a hard physical deadline: the heralding photon’s classical confirmation takes about 1 millisecond round trip, so the memory must hold entanglement, at high retrieval efficiency, for many milliseconds to be useful — and repeat that across multiple swap rounds. Ensembles decay too fast, long-lived spins herald too rarely, and the 420-km record explicitly sidestepped the requirement by reading out before the herald returned.

Established Quantum frequency conversion currently costs 88 to 95 percent of every photon. Harvard’s 737-to-1,350 nm converter ran at 5.4 percent total efficiency; the Hefei ion node’s at 12 percent. Since a two-segment swap needs conversion at both ends of both segments, present converters alone impose a rate penalty of order one thousand. Native-telecom emitters (erbium, specific quantum dots) or vapour memories with near-telecom lines (Qunnect’s 1,324 nm) trade this cost against other corners of the triangle.

Established Deployed fibre is a hostile instrument: phase drift, polarisation wander, and thermal transients that laboratory spools do not have. Twin-field QKD and single-photon-interference entanglement schemes require subwavelength path-length stability over hundreds of kilometres; the 420-km experiment needed a dedicated dual-band locking scheme with a separate 1,600 nm reference laser, and Toshiba’s 254-km link is notable mostly for solving exactly this on live commercial fibre. Qunnect’s 15-day uptime shows polarisation drift is automatable at metropolitan scale; nobody has shown the same at 500 km.

Established Detector infrastructure is cryogenic. The rates quoted throughout this brief lean on superconducting nanowire single-photon detectors at 90-plus percent efficiency, which need closed-cycle cryostats at 1–3 K in every heralding station; diamond and silicon-vacancy nodes additionally need dilution refrigerators below 200 mK. Toshiba’s avalanche-photodiode result shows QKD can shed the cryostat; no memory platform yet can.

Frontier Rate arithmetic remains brutal even after multiplexing. The best memory-to-memory rate over metropolitan fibre is a few per second; entanglement purification — needed to lift 78–90 percent fidelities toward the 99-plus percent that error-corrected applications require — consumes multiple raw pairs per purified pair, and swapping multiplies losses. Realistic repeater simulations demand thousands of high-fidelity pairs per second per segment; the field is three to four orders of magnitude short.

Frontier There is no certification regime. Security agencies’ core complaint about QKD — that implementation security cannot yet be evaluated the way algorithmic cryptography can — applies with more force to entanglement networks, for which no evaluation methodology, protection profile, or interoperability test suite exists at all. A network nobody can accredit is a network critical-infrastructure operators cannot legally buy.

Established Progress rides on five fields the network community mostly does not fund. Atomic, molecular and optical physics supplies every memory platform; integrated photonics supplies the cavities and modulators that raised heralding efficiencies past 90 percent in diamond nanostructures; cryogenic engineering sets the cost floor of every node; ultralow-loss fibre manufacture (0.157–0.17 dB per km in current G.654-class glass) sets the loss budget; and time-frequency metrology supplies the phase references and White-Rabbit-class synchronisation every heralded scheme assumes.

Frontier Hollow-core fibre could move the loss floor itself. Attenuation below 0.1 dB per km was reported in laboratory spans in 2024; if it survives cabling and deployment, the exponential that defines the field softens for everyone at once — a rare dependency that helps quantum and classical traffic equally.

Established Post-quantum cryptography is a dependency in the negative: it removes the field’s most marketable justification. With NIST-standardised lattice cryptography shipping in commodity software since 2024, quantum networking’s case must rest on what entanglement alone provides — distributed computing, device-independent security, and nonlocal sensing — not on key exchange. Roadmaps that still lead with QKD are, on the agencies’ published record, leading with their weakest card.

Established Theory is ahead of hardware and occasionally load-bearing. The PLOB bound defines what “beating direct transmission” means; entanglement distillation and swapping protocols date to the 1990s; RFC 9340 and the simulation stacks at Delft, Chicago and Hefei define the control planes hardware will eventually meet. Nothing on the theory side blocks a repeater; everything on the materials side does.

Frontier Satellite links are the complement, not the competitor, of fibre. Micius proved entanglement survives 1,200 km of vacuum and atmosphere with total loss far below fibre’s at that distance; a hybrid architecture — satellites for the long haul, fibre and memories for the metro — depends on daylight-capable optical terminals and on memories that can buffer a pass, neither demonstrated together.

Frontier The decisive demonstration is a memory-enhanced repeater link over deployed fibre, between independently operated nodes in different buildings, that delivers entangled pairs at a higher rate than direct transmission through the same fibre would allow. Nobody has done it; every credible roadmap treats it as the gate through which the field must pass. The components have each been shown: repeater-advantage in one lab (Harvard 2020), metropolitan memory links in three countries (2024–2026), beyond-repeaterless scaling at 420 km of loss (2026). Groups in Hefei, Delft and Boston are explicitly building toward the assembled version, and it is plausibly a this-decade result; until it lands, every quantum-internet roadmap rests on extrapolation.

Frontier Second: an entanglement swap across two deployed segments, joining end nodes in different cities that share no direct fibre, with an end-to-end Bell violation. This is the smallest experiment that is a network rather than a link — three sites, two swaps of trust into physics — and it would retire the routing question from simulation to measurement. The Hefei three-node system performs the topology inside one metropolitan area; no two-city version exists.

Frontier Third: entangle a superconducting processor qubit with a telecom photon across more than 1 km of fibre at fidelity above one-half. This is the transduction milestone that decides whether the dominant computing platform can ever be networked directly, or whether distributed quantum computing waits for optically native processors.

Established Fourth: an interoperability test — two independently built nodes, different vendors or laboratories, establishing entanglement through a standardised heralding interface. Every demonstration to date is single-team hardware at both ends. Classical networking became an industry the day alien routers interoperated; the quantum equivalent has never been attempted and requires no new physics, only agreed interfaces — which is why its absence is an institutional datum, not a technical one.

Speculative Fifth: an application result — a clock comparison, baseline interferometry run, or verified blind computation over a deployed entanglement link that beats the best classical method at the same task. No such result exists on any testbed. One would convert the field’s economics overnight, which is precisely why its absence after two decades of testbeds deserves flagging rather than excuse.

Established The engineering frontier is making quantum hardware live in telecom facilities on telecom terms. Toshiba’s 254-km system ran in standard racks with room-temperature detectors on commercially operated fibre; that — not the key rate — is its significance. Every memory platform still fails this test: diamond nodes need dilution refrigerators, ion nodes need optical tables and vacuum systems, ensemble nodes need laser cooling. Qunnect’s warm-vapour line is the deliberate exception — rack-mounted, cryogen-free — bought at the price of microsecond storage.

Established Coexistence with classical traffic is proven at metro scale and unproven beyond it. Quantum channels at single-photon power levels must share conduits — and ideally fibres — with classical signals twelve orders of magnitude brighter; O-band quantum alongside C-band data with careful spectral filtering has worked in New York and Boston deployments. Dedicated dark fibre remains the default for every record result, and dark fibre at continental scale is a supply-chain item, not a catalogue item.

Established Automation is the difference between an experiment and infrastructure, and it has been demonstrated exactly once at length: 15 days unattended at 99.84 percent uptime on the New York link. Every memory-based network still runs with physicists in the loop. Heralding stations need remote calibration, drift compensation, and failover; none of this is exotic, and almost none of it is done.

Frontier Mid-span stations are a civil-engineering fact the diagrams hide. Single-photon-interference schemes put the beamsplitter and detectors at the fibre midpoint — XingHan 2.0’s and the 420-km experiment’s middle node “Charlie” — so a repeater backbone means conditioned, powered, cryostat-capable huts every 50–100 km, on rights-of-way someone must lease. The cost model looks like cellular backhaul, not like software.

Frontier The control plane must treat entanglement as perishable inventory. Pairs expire in milliseconds; scheduling, purification and swapping decisions happen faster than wide-area round trips, which pushes toward local autonomy at nodes and pre-agreed policy — an architecture RFC 9340 sketches and no deployment has stress-tested.

Established Post-quantum cryptography is the adjacent technology that already ate the first market. Lattice-based key exchange standardised by NIST in 2024 does in software most of what trusted-node QKD does in hardware, which is why this brief treats key distribution as quantum networking’s past and entanglement services as its only future.

Established Superconducting detector and cryogenic supply chains are shared with the computing and astronomy worlds. Nanowire detectors and millikelvin platforms tie this field’s cost curve to progress covered in High Temperature Superconductors and Superconducting Infrastructure; a detector that worked at 77 K would change every node’s economics.

Frontier Free-space optical links are converging with this field from two directions. Satellite entanglement distribution (Micius) shares terminals, adaptive optics and pointing with the deep-space optical communication programme described in Deep Space Communications; a memory-buffered satellite pass is the only credible intercontinental architecture before repeater chains mature.

Frontier Networked sensing is the application adjacency with the earliest plausible customers. Entanglement-linked optical clocks would sharpen time transfer and geodesy, and proposed entanglement-assisted very-long-baseline interferometry would effectively extend the apertures discussed in Mega Telescopes; both remain paper protocols awaiting the memory rates this brief documents as missing.

Speculative Distributed and blind quantum computing inherit whatever the network achieves. Modular quantum computers linked by entanglement are the standard answer to monolithic scaling limits, and verifiable delegated computation is cryptographically attractive; both assume per-link entanglement rates and fidelities three orders of magnitude beyond the measured record, which is why they sit in the adjacency list and not the position section.

Established The field’s geography is three national programmes and their telecom partners. China’s effort runs through USTC and the Hefei National Laboratory with fibre from China Unicom and dedicated vendors (QuantumCTek), and holds most measured records. The EU funds the Quantum Flagship and the EuroQCI initiative — all 27 member states committed to national quantum-communication segments, several hundred million euros programmed — with the Quantum Internet Alliance targeting prototype intercity entanglement networks around the turn of the decade (programme’s own goals). The United States funds through the National Quantum Initiative and the Department of Energy’s 2020 quantum-internet blueprint, with the Chicago Quantum Exchange’s roughly 200-km metropolitan testbed linking Argonne, Fermilab and campus sites, and the Boston and New York links supplying the deployed-fibre results above.

Established Standards exist for the technology the agencies do not want, and barely exist for the one everyone says is the goal. ETSI’s QKD industry specification group has published interface and security specifications since 2008; the ITU-T Y.3800 series standardises QKD network layering; and the IETF’s RFC 9340 (2023) is an informational architecture document for a quantum internet with no protocol standards beneath it. There is no interoperability standard, no conformance suite, and no certification scheme for entanglement-based networking anywhere.

Established Procurement is split down the middle of Western governments. Security agencies (NSA, NCSC, BSI, ANSSI) advise against QKD purchases while innovation and infrastructure ministries fund QKD deployment through EuroQCI and national testbeds — a standing contradiction sustained because the funding streams answer to different departments. Telecom operators hedge: Deutsche Telekom hosts Toshiba’s flagship link, BT ran a commercial London metro QKD service with Toshiba from 2022 (vendor deployments).

Frontier The institutional gap that matters is an accreditation pathway for entanglement-based security. Device-independent protocols are, on paper, exactly what certification-averse agencies say they want — security auditable from measured Bell statistics rather than trusted hardware — yet no agency has published evaluation criteria for them, and no lab could presently submit a product. Whoever builds that pathway first sets the terms of the eventual market.

Established The live ethical issue in 2026 is honesty in threat marketing, not misuse of the technology. The harvest-now-decrypt-later threat is real — encrypted intercepts stored today become readable if cryptanalytically relevant quantum computers arrive — but the proportionate answer, per every Western security agency on record, is software migration to post-quantum cryptography, already standardised and shipping. Selling nation-scale QKD hardware against that threat, at hundreds of millions of euros of public money, is a policy choice its own security services have publicly declined to endorse; the public record deserves to say so plainly.

Frontier Entanglement networks redistribute trust rather than creating surveillance capacity. A functioning quantum network removes the trusted middleman — the relay operator who could read every key — and device-independent protocols would let users verify security without trusting vendors or certifiers. The civil-liberties valence of that is broadly positive; the losers are lawful-intercept architectures, and the tension with interception mandates in several jurisdictions has received almost no policy attention.

Speculative Access asymmetry is the plausible long-run concern. If entanglement-secured or entanglement-enhanced services mature, early capacity will concentrate in three blocs and their metropolitan cores; a security or sensing premium available only on subsidised backbones would repeat familiar infrastructure inequities — a concern whose scale depends entirely on applications that do not yet exist.

Established Research-integrity pressure is measurable in this field. Distance and rate records are routinely reported in headlines without the spooled-versus-deployed, post-selected-versus-heralded distinctions that determine what was demonstrated; the 420-km case, where independent analysis had to surface that both nodes shared one laboratory, is a working example of why this brief marks configurations explicitly.

Speculative If repeater networks mature, the durable civilizational payload is distributed quantum computation and verifiable computing, not secrecy. Modular machines linked by entanglement are the standard route past single-cryostat scaling limits, and blind quantum computing would let any client verifiably delegate computation to an untrusted server — a structural change in who must be trusted with what, comparable to what public-key cryptography did for commerce.

Speculative A global entangled clock and sensor fabric would quietly upgrade the measurement substrate of civilisation. Continental clock comparison at the 10^-18 level bears on geodesy, navigation independence, and tests of general relativity; entanglement-extended interferometric baselines would sharpen astronomy; both connect this brief to the experimental programmes described in Quantum Gravity, where networked quantum systems are among the few proposed instruments for probing gravity’s quantum character.

Established The quantum internet, if it arrives, will be an overlay for special tasks, not a successor to the internet. Every quantum protocol consumes classical communication; entanglement carries no controllable information by itself; and nothing about the technology moves bulk data. The correct mental model is GPS to roads: a thin, precise service layered on massive classical infrastructure.

Handwave Claims that quantum networks are a prerequisite for interstellar-grade civilisation, or for fundamentally unbreakable global order, work by assertion. No articulated mechanism connects entanglement distribution to either outcome; the honest long-range statement is that a civilisation with mature quantum networks has better clocks, better telescopes, provably private computation, and no faster messages than light allows.

These horizons track entanglement networking — trusted-node QKD is excluded as already deployed and already declining as a policy preference.

• 10 yr: Frontier A field repeater passes break-even on at least one deployed metropolitan link; two-city entanglement swapping with end-to-end Bell certification; automated multi-node metro networks in Hefei, the Netherlands, Chicago and Boston running continuously; first vendor-to-vendor interoperability demonstration; transduction reaches tens of percent in the lab. Key rates and memory lifetimes each gain one order of magnitude.
• 25 yr: Speculative Intercity repeater chains at 300–800 km connecting national metros; hybrid satellite-plus-memory links spanning continents; early distributed-computing and clock-network services with paying institutional users, if and only if the break-even repeater arrived on the 10-year horizon.
• 50 yr: Speculative Continental entanglement backbones with error-corrected end-to-end links; device-independent security products with an accreditation regime; entanglement-assisted baselines in operational astronomy and geodesy.
• 100 / 250+ yr: Handwave Solar-system-scale entanglement distribution riding the deep-space optical infrastructure, entangled observatories at interplanetary baselines, and quantum-networked computing as an assumed utility — extrapolations with no anchoring demonstration at any scale today.

• Depends on No brief on this map blocks metropolitan entanglement networks — the binding constraints are laboratory-scale physics results. Node economics inherit from progress in High Temperature Superconductors (detector and cryostat cost) and Superconducting Infrastructure; intercontinental reach inherits optical-terminal work covered in Deep Space Communications.
• Requires (not on this map) A quantum memory combining second-scale storage, high retrieval efficiency and telecom-wavelength operation, which no platform’s triangle yet closes; microwave-to-optical transduction near unit efficiency at single-photon noise levels so superconducting processors can join fibre networks; cryogen-free rack-mounted network nodes manufacturable at telecom-equipment cost rather than physics-laboratory cost; continental dark-fibre access with powered mid-span stations every 50–100 km for heralding hardware; a security-accreditation pathway under which agencies now hostile to QKD could certify entanglement-based products; and a paying application for distributed entanglement beyond key exchange, without which the market case collapses onto ground the security agencies have already contested.
• Enables Modular distributed quantum computing, blind and verifiable delegated computation, device-independent cryptography, entangled clock networks for geodesy and time transfer, and entanglement-extended sensing baselines.
• Adjacent Post-quantum cryptography as the software competitor for the security market; satellite optical links shared with Deep Space Communications; instrument-scale payoffs for Mega Telescopes; and networked tests bearing on Quantum Gravity.

Handwave “Entanglement enables faster-than-light communication.” It does not, and this is a theorem, not an engineering gap: measurement outcomes on entangled particles are locally random, and the no-signalling theorem guarantees no controllable information crosses without a classical channel, which travels at light speed or slower. Every quantum-network protocol explicitly consumes classical communication; teleportation moves a quantum state only after classical bits arrive.

Established “China already has a quantum internet.” China operates the world’s largest quantum-secured key network — 4,600 km, trusted relays throughout — and holds most entanglement records, but a trusted-relay chain is stage zero on the field’s own ladder: keys sit unencrypted at every relay, and no entanglement crosses any node. The claim conflates the two ledgers this brief keeps apart.

Established “QKD is unhackable because physics.” The theorems secure ideal devices; real devices have been breached repeatedly through implementation attacks (detector blinding demonstrated against commercial systems in 2010, with a family of successors), and the impossibility of certifying implementations is the stated core of the NSA, NCSC and joint European agency positions declining to endorse QKD. “Physically guaranteed security” in vendor material describes the model, not the product.

Established “The 420-km result means intercity quantum repeaters exist.” The 2026 record entangled two memories in the same laboratory through mostly spooled fibre, with readout before the herald returned; the largest geographic separation of entangled memories anywhere remains 14.5 km (Hefei) and roughly 10 km between Delft and The Hague. Channel-loss benchmarks and networks are different achievements, and the field’s own papers are careful about this even when coverage is not.

Handwave “A quantum internet will replace today’s internet.” Nothing in the technology moves bulk data; entanglement is a consumable resource for specific protocols, produced today at per-second rates against classical links carrying terabits per second. The quantum internet is an overlay service, and every serious architecture document (RFC 9340 included) says so.

Frontier “Post-quantum cryptography makes quantum networks pointless.” Half true, and the half matters: PQC removes the key-distribution rationale — the agencies are explicit — but it does nothing for distributed quantum computing, blind computation, entangled sensing, or device-independent protocols, which require entanglement as a physical resource no algorithm substitutes. The honest restatement is that PQC killed the weak case for quantum networks and left the strong, unproven one.

Frontier “Room-temperature quantum memories have solved deployment.” Warm-vapour memories (86.5 percent entanglement fidelity at 1,200 pairs per second, telecom-native, in a 2025 vendor preprint) genuinely remove cryogenics — and store for 3 microseconds, about a thousandfold short of a 100-km repeater segment’s round-trip requirement. One corner of the triangle moved; the triangle remains.

Speculative “A continental quantum internet by 2030.” Programme roadmaps (EU flagship documents among them) carry dates in this range for prototype intercity networks; on the measured record — no field repeater, rates of a few pairs per second, no interoperability standard — prototype intercity links by 2030 are plausible, continental service is not, and any roadmap that does not distinguish the two is marketing.