Hadamard Get the app

How Far Can Quantum Key Distribution Reach? 1,002 km, With Caveats

How Far Can Quantum Key Distribution Reach? 1,002 km, With Caveats — Hadamard

How far can quantum key distribution reach? The fiber record without any relay is 1,002 km, and satellites have connected ground stations about 12,900 km apart. The catch is that the fiber record produces a key rate close to zero, and every intercontinental link so far has depended on a satellite acting as a trusted relay rather than on end-to-end quantum security.

This article covers the record, the physics behind the limit, what “intercontinental QKD” means in practice, and the 2026 results that point past trusted relays.

How far can quantum key distribution reach? The short answer

The numbers depend on what you count as “reach.” The table below lists the main reference points from the published literature.

Link Distance Type Reported key yield End-to-end quantum security?
USTC twin-field QKD (2023) 1,002 km fiber Point-to-point, no relay 0.0034 bps Yes, between the two ends
Micius, China–Austria (2018) 7,600 km Satellite Roughly kHz satellite-to-ground rates per pass No, satellite is a trusted relay
Jinan-1, Beijing–Stellenbosch (2025) About 12,900 km Microsatellite Up to 1 Mbit of secure key per pass No, satellite is a trusted relay
Linköping–Stockholm (2026) 303 km fiber Terrestrial trusted node Not covered here No, intermediate node is trusted
USTC device-independent QKD (2026) 11 km practical, 100 km feasibility test Single-atom nodes Not covered here Yes, devices need not be trusted

Treat the satellite and the fiber numbers as different kinds of result. The fiber record is a single quantum channel. The satellite figures describe two separate quantum links joined by a node that holds both keys.

Why distance kills QKD, and what the PLOB bound says

No amplifiers, exponential loss

Classical fiber networks cross oceans by amplifying the signal every so often. QKD can’t do that. In the abstract of the 1,002 km paper, the authors state the problem plainly: quantum signals cannot be amplified, and channel loss grows exponentially with distance. The no-amplification rule is what makes QKD secure, because an eavesdropper can’t copy the photons either.

Fiber loss is quoted in decibels per kilometer, so the loss in decibels grows in step with distance while the fraction of photons that survive shrinks exponentially. The 1,002 km experiment used ultra-low-loss pure-silica-core fiber with a maximum attenuation of 0.16 dB/km. Our own arithmetic: at that maximum, 1,002 km comes to roughly 160 dB of loss. That means about one photon in 1016 gets through.

The PLOB bound: the ceiling for any repeaterless link

In 2015, Pirandola, Laurenza, Ottaviani and Banchi worked out exactly how much secret key a lossy channel can carry. The paper was published in Nature Communications in 2017. For a lossy channel with transmissivity η, they showed that “Q2=K=−log2(1−η)”. In the high-loss regime they found “the optimal rate-loss scaling of K≃1.44η secret bits per channel use.”

This is now known as the PLOB bound. The authors describe it as “setting the ultimate and precise limit that only quantum repeaters can surpass” (full text on ar5iv). The practical point is that the key rate falls linearly with η, and η itself falls exponentially with distance. A direct point-to-point link has no way around this.

Twin-field QKD: how the 1,002 km fiber record was set

Twin-field QKD was proposed by Lucamarini, Yuan, Dynes and Shields at Toshiba in a 2018 Nature paper titled “Overcoming the rate-distance barrier of quantum key distribution without using quantum repeaters.” In their scheme the key rate scales with the square root of the channel transmittance, as it would with a quantum repeater, but no repeater is needed. They also reported that it was feasible with current technology and that noise stayed manageable over 550 km of standard fiber.

The scaling difference is large. If η is 10−16, √η is 10−8, eight orders of magnitude higher. According to the USTC team, its sending-or-not-sending (SNS) variant changes the key rate’s dependence on transmittance from η to √η.

The record result, “Experimental Twin-Field Quantum Key Distribution over 1000 km Fiber Distance,” appeared in Physical Review Letters on 25 May 2023. The authors include Xiang-Bin Wang, Qiang Zhang and Jian-Wei Pan. Its main ingredients were:

  • 3-intensity SNS TF-QKD with actively-odd-parity pairing
  • Dual-band phase estimation
  • Ultra-low-noise superconducting nanowire single-photon detectors, with system noise of about 0.02 Hz according to the paper
  • Detectors from SIMIT filtered at 40 K and 2.2 K, with detector noise of about 0.02 cps according to the USTC press page
  • Low-loss fiber made with YOFC

The paper reports a key rate of 9.53×10−12 per pulse over 1,002 km in the asymptotic limit. With finite-size effects included, it reports 8.75×10−12 per pulse at 952 km. A follow-up paper with full finite-key analysis reports a secure key over 1,002 km at 3.11×10−12 per pulse. The same follow-up reports 111.74 kbps at 202 km.

For a rough comparison, our own arithmetic again: at about 160 dB of loss, the PLOB linear scaling gives around 1.44×10−16 bits per channel use. The reported per-pulse rates are several orders of magnitude higher, and that gap is exactly what √η scaling is meant to deliver.

What 0.0034 bits per second means in practice

USTC describes the result as a new world record for non-relay QKD, with a key rate of 0.0034 bps. A simple calculation shows what that means: one 256-bit symmetric key would take around 75,000 seconds to build up, which is roughly 21 hours.

That makes the 1,002 km result a physics milestone, not a usable range. The figures at 202 km show where useful rates are found today. If someone asks how far a QKD link can reach and still be practical, the answer is a few hundred kilometers of fiber, not a thousand.

Satellite QKD distance: Micius and Jinan-1 as trusted relays

Satellites avoid most of the fiber problem because much of the photon path runs through vacuum. Both of the record-length satellite links, however, are built as two separate keys joined by a relay.

Micius: China to Austria, 7,600 km

A 2018 Physical Review Letters paper with 36 authors, among them Sheng-Kai Liao, Anton Zeilinger and Jian-Wei Pan, reported secret keys shared between China and Europe at locations 7,600 km apart. Satellite-to-ground keys were generated at roughly kHz rates per pass.

The paper is explicit about how the link worked. Micius shared one key with a Chinese ground station and another with an Austrian one. On command, it XORed the two keys and sent the result to one of the stations, which then had what it needed to recover the other station’s key. The satellite therefore held both keys and served as a trusted relay.

Jinan-1: Beijing to Stellenbosch, about 12,900 km

In March 2025, a USTC team including Pan Jianwei, Peng Chengzhi and Liao Shengkai published real-time QKD from the Jinan-1 microsatellite in Nature. According to the EurekAlert release, the system sent about 250 million quantum photons per second and produced up to 1 Mbit of secure key per pass. The ground stations weighed about 100 kg, two orders of magnitude lighter than before.

Ground stations were set up in Jinan, Hefei, Nanshan, Wuhan, Beijing, Shanghai and Stellenbosch in South Africa. Keys and encrypted communication were exchanged between Beijing and Stellenbosch, about 12,900 km apart, with the satellite again acting as a trusted relay. A Nature reviewer called the work “a milestone for the realization of a satellite constellation for quantum and classical communication.”

Europe’s planned QKD satellite has slipped, as covered in our article on why the Eagle-1 launch date moved to 2027–28.

Trusted node QKD: the 303 km Swedish deployment

Long QKD links on the ground work the same way: they are chained from shorter links. A June 2026 preprint by Clason, Argillander, Xavier and colleagues describes a 303 km trusted-node link between Linköping University and Stockholm. It consists of 270 km of deployed single-mode fiber and a 33 km multi-core fiber segment. The link used commercial QKD systems with external superconducting detectors, and the keys were used for one-time-pad image encryption.

Trusted nodes are a reasonable engineering compromise, but the security model changes. Each node holds the key in usable form, so the link is only as secure as the least protected building in the chain. That brings back the kind of infrastructure trust QKD was supposed to remove, which is one reason security agencies remain cautious (see our article on the NSA’s 2026 position on QKD).

Beyond trusted relays: USTC’s 2026 repeater and DI-QKD results

The PLOB authors pointed to the long-term fix: quantum repeaters, which pass entanglement along a chain without ever measuring or storing the key in the middle. In early 2026, a USTC team led by Pan Jianwei, Zhang Qiang and Bao Xiaohui published two related results, according to the Chinese Academy of Sciences:

  • Nature, 2 January 2026: “Long-lived remote ion-ion entanglement for scalable quantum repeaters.” The work showed memory-memory entanglement between two nodes joined by entanglement swapping.
  • Science, 5 January 2026: “Device-independent quantum key distribution over 100 km with single atoms.”

Device-independent QKD gets its security from a Bell inequality violation, so users don’t need to trust their own hardware. It uses the same idea as the E91 protocol and its Bell tests. Physics World reports that earlier DI-QKD had reached about 700 m at best. The CAS release says it had been limited to a few hundred meters.

The details matter here. According to Physics World, practical DI-QKD with finite-key security was shown over 11 km of fiber, and the 100 km figure was a feasibility test only. The nodes were single trapped atoms, with 780 nm photons converted to 1315 nm. Physicist Carlos Sabín, who was not involved, noted error rates of about 3% at 11 km, rising above 7% at 100 km, and said “there is still a long way to go.” Pan’s team says: “We are aiming for 1000 km,” using quantum repeaters.

For background on how repeaters and entanglement swapping fit together, see our explainer on quantum internet repeaters and real links.

What this means for secure communications today

For anyone planning real systems, the published record supports a few conclusions:

  1. Useful point-to-point QKD is a metro and regional technology. Rates in the kbps range have been reported at around 200 km. At 1,000 km the rate is a fraction of a bit per second.
  2. “Intercontinental QKD” means trusted relays. Both Micius and Jinan-1 relayed keys through a satellite that held them.
  3. End-to-end long-distance QKD needs quantum repeaters. The 2026 results are real progress, but practical DI-QKD currently reaches 11 km.
  4. QKD needs dedicated hardware on every link. That includes low-loss fiber or optical ground stations, plus superconducting detectors in the record experiments.

Most systems that need quantum resistance today use software instead, through post-quantum algorithms on existing networks. Hadamard takes that approach with BB84, an Ethereum Layer 2 that uses hash-based one-time addresses instead of ECDSA keys. Despite the shared name, it does not use the BB84 QKD protocol or any quantum hardware.

FAQ

What is the longest distance quantum key distribution has worked over fiber?

The longest is 1,002 km, using twin-field QKD, published in Physical Review Letters in May 2023 by a USTC-led team. USTC calls it the world record for non-relay QKD, with a key rate of 0.0034 bps.

Did China really send quantum keys 12,900 km, and is that link end-to-end secure?

Keys and encrypted communication were exchanged between Beijing and Stellenbosch, about 12,900 km apart, using the Jinan-1 microsatellite. The satellite served as a trusted relay, so the link is not end-to-end quantum secure. Anyone who could compromise the satellite could get the key.

Why can’t QKD signals just be amplified like normal fiber traffic?

Quantum signals can’t be amplified without disturbing them, and that same property is what stops an eavesdropper from copying them. Loss therefore builds up exponentially with distance, and the PLOB bound caps the rate of any link without a repeater.

When will quantum repeaters make long-distance QKD practical without trusted nodes?

No timeline has been established. In 2026, USTC demonstrated memory-memory entanglement between two repeater nodes and practical device-independent QKD over 11 km, and the team says it is aiming for 1,000 km. Outside researchers say there is still a long way to go.

Sources