Applications of the quantum internet

Quantum key distribution (QKD)

QKD is the only application of quantum networking with commercial deployment today. Two paradigm families: prepare-and-measure protocols (BB84 and its variants, with decoy states for security against photon-number-splitting attacks) where Alice prepares states and Bob measures; and entanglement-based protocols (BBM92, Ekert91) where a shared Bell pair is the resource. Variants address specific operational constraints: Measurement-Device-Independent QKD (MDI-QKD) puts the trust in the source and removes detector side-channel attacks; Twin-Field QKD (TF-QKD) scales the key rate as √η rather than η, doubling effective reach, and now holds the fibre-distance record at 1002 km ^{Liu et al. 2023} . A separate research thread, High-Dimensional QKD (HD-QKD), encodes each photon in a qudit (d > 2, via time-bin or orbital-angular-momentum modes) to raise bits-per-photon and noise tolerance ^{Islam et al. 2017} — still academic; no commercial deployments.

QKD requires the lowest fidelity threshold of any quantum-network application — ~0.85 raw is workable after error correction and privacy amplification, and field-deployed systems run reliably between metropolitan-scale endpoints. The bulk of vendor activity in quantum networking is QKD ^{Kumar et al. 2025} .

Distributed quantum computing (DQC)

A single quantum processor with a fixed qubit count is bounded by what its chip-scale platform allows; networking processors into a cluster lets the combined machine run circuits larger than any single node. The networking primitive is gate teleportation: a CNOT between qubits at two different nodes is performed by consuming a Bell pair and exchanging a small number of classical bits. The protocol generalises to other two-qubit gates and to higher-fidelity primitives via purification.

DQC's demands on the network are strict: gate teleportation requires fidelities above ~0.99 to compose into a useful multi-gate circuit. The end-to-end fidelity drops multiplicatively with depth, so DQC sits at the high-fidelity end of the application spectrum and depends critically on purification ^{Kumar et al. 2025} .

Blind quantum computing (BQC)

A client has sensitive input data and wants a remote quantum server to compute on it without learning the inputs or outputs. Blind quantum computing protocols achieve this: the client prepares (or has delivered) qubits in known states known only to them, sends them to the server, and the server executes the computation through measurement-based protocols on those qubits. Because the server never learns the state encoding, it learns nothing about the data.

BQC is appealing for high-value privacy use cases — financial computations, pharmaceutical search, government applications. Like DQC it requires high fidelity and is currently research-stage.

Networked sensing & clock synchronisation

Entanglement between sensors at different locations improves measurement precision beyond what classical correlation allows. Two well-developed applications:

• Networked atomic clocks. Distributing entangled states between geographically separated atomic clocks lets the network's frequency measurement scale better than √N with the number of clocks (Heisenberg-limited rather than shot-noise-limited).
• Distributed interferometry. Very-long-baseline interferometry (VLBI) for astronomy can in principle use entangled photon pairs to extend the effective baseline. This is research-level but theoretically promising.

Sensing applications consume entanglement differently from QKD or DQC — the fidelity threshold can be more relaxed in some protocols, but the source-of-randomness and timing-precision constraints are tighter ^{Kumar et al. 2025} .

Quantum position verification

A verifier wants to confirm that a prover is physically located at a claimed point in space — not impersonated by a relay, not co-located with an attacker, not anywhere else on the network. The classical construction is to send a challenge from several reference stations and check that the response arrives with the round-trip timing consistent with the speed of light from the claimed position. Chandran, Goyal, Moriarty and Ostrovsky showed in 2009 that this classical scheme is insecure against colluding adversaries who can record and forward each challenge in parallel ^{Chandran et al. 2014} . Replacing the classical challenge with a quantum state forces a different attack model: the no-cloning theorem prevents simple copy-and-forward, so quantum position verification (QPV) is a primitive native to a network that delivers quantum states, not just keys.

The catch came quickly. Buhrman, Chandran, Fehr, Gelles, Goyal, Ostrovsky and Schaffner proved that for any QPV protocol, a coalition of adversaries surrounding the claimed position can fake the prover if they share enough pre-distributed entanglement — the attack reduces to instantaneous non-local quantum computation, which is unconditionally possible given unbounded entanglement ^{Buhrman et al. 2014} . Earlier teleportation-based attacks on specific proposals were already known from Kent, Munro and Spiller ^{Kent et al. 2011} . The line of research that survives is protocols where the attackers' entanglement cost grows with the security parameter, so a bounded adversary is provably defeated.

Two protocol families dominate the constructive direction. f-routing asks the prover to route a quantum state to one of two verifiers depending on a classical bit. BB84-style QPV sends a single BB84 qubit alongside a classical basis announcement and requires the prover to measure and return the outcome to both verifiers. Bluhm, Christandl and Speelman constructed an interleaved-product variant whose best known attack consumes entanglement exponential in the number of classical bits exchanged, using only a single qubit on the prover's side ^{Bluhm et al. 2022} . Loss in the verifier-to-prover channel was a security loophole until Allerstorfer et al. (2025) showed that adding a photon-presence pre-commitment step reduces the security-relevant loss to that of the prover's local apparatus, restoring the original attack bounds for BB84-style QPV ^{Allerstorfer et al. 2025} . Experimental work is early: Kanneworff et al. (2025) demonstrated the optical building blocks of a loss-tolerant scheme with a true single-photon source, but a full end-to-end QPV deployment with spatially separated verifiers has not been reported ^{Kanneworff et al. 2025} .

QPV binds an identity claim to a physical location in a way no purely classical or PQC primitive can. The use cases are geo-locking of compute resources to named data centres, defence against man-in-the-middle relays where the classical control plane is untrusted, and verifiable delivery of quantum state to a known endpoint. It sits beyond stage 2 of the Wehner roadmap — it needs the channel to carry quantum states, not just to extract keys from them.

Multiparty quantum consensus

Byzantine agreement is the textbook distributed-consensus problem: $n$ parties, of which up to $f$ may be arbitrarily faulty, must reach a common decision. Classically the problem is unsolvable when $n<3f+1$ — the three-party case with one faulty party is provably impossible. Fitzi, Gisin, and Maurer (2001) showed that detectable broadcast, the natural quantum relaxation, is achievable in exactly this three-party regime using pairwise quantum channels and entangled qutrits ^{Fitzi et al. 2001} .

Gaertner et al. (2008) realised the protocol experimentally with a four-photon entangled state, the first demonstration that the quantum advantage is not purely theoretical ^{Gaertner et al. 2008} . The current experimental state of the art is Jing et al. (2024) — a three-user quantum-network demonstration on integrated photonics that achieves Byzantine agreement with unconditional security beyond the classical fault-tolerance bound ^{Jing et al. 2024} .

What consensus needs from the network: multi-party entanglement (four-photon GHZ-style states in the worst case) delivered with high fidelity between three or more endpoints, plus a classical side-channel for the broadcast phase. Rate requirements are modest — agreement protocols run once per decision, not continuously — but fidelity is high (the security argument tightens as raw fidelity rises) and the network has to support simultaneous delivery to multiple endpoints, not just point-to-point. Application areas include permissioned blockchains, distributed leader election, and any setting where a classical majority assumption is unavailable.

Cross-application demands

The six application families have qualitatively different requirements on rate, fidelity, and reach:

The metrics subject covers what these demands mean for a network's engineering budget; the maturity subject covers where each application sits on the TRL ladder ^{Cacciapuoti et al. 2020} .