Quantum Use Cases in Telecom

The telecommunications industry has become one of the most active buyers of quantum promises. Operators on four continents are running pilot programmes, signing vendor partnerships, and briefing their boards on quantum readiness. What they are getting for that investment varies enormously: from production-deployed atomic clocks that solve a real operational problem today, to optimization demos that compare a quantum annealer against an unnamed “general-purpose solver” and declare victory.

I have spent the past several months assessing these programmes, and I separate the findings into four fronts: quantum computing, quantum annealing and Ising machines, quantum sensing and timing, and quantum communications. The four differ in demonstrated results, timeline, and claim-evidence gap. A CISO or CTO making investment decisions needs them separated: what should my organization do this year?

The bottom line: the nearest-term, procurable-today value for a telco is in timing and synchronization (caesium-disciplined enhanced primary reference time clocks) and quantum random number generation (QRNG). Quantum-inspired classical Ising machines can solve real combinatorial problems in network planning; telcos should benchmark them against existing tools. Everything else (gate-model quantum computing for optimization, quantum annealing advantage over classical solvers, quantum key distribution as a standalone security architecture) either lacks demonstrated advantage or has unresolved limitations that the marketing materials do not mention.

Quantum computing for telecom optimization

Telecom networks generate hard combinatorial problems. Cell site placement, antenna tilt and power configuration, spectrum allocation, routing and wavelength assignment (RWA), network slicing, VNF scheduling, workforce dispatch: all are combinatorially complex, and operators have been told that quantum computers can solve them better than classical methods.

Quantum annealers and the Quantum Approximate Optimization Algorithm (QAOA) can accept most of these problems as QUBO (quadratic unconstrained binary optimization) or Ising inputs. Some are mixed-integer linear programmes (MILP) suited to commercial solvers like Gurobi and CPLEX. A few are graph problems. The formulations exist and the mappings are published. Macaluso, Geraci, Combarro and colleagues surveyed the landscape in IEEE Communications Magazine (published online August 2025, January 2026 issue). The survey maps telecom problems to quantum formulations and concedes that no demonstrated advantage exists at scale.

The results, however, do not support the claims.

Not a single telecom quantum-computing use case has demonstrated benchmarked quantum advantage over well-tuned classical methods on the same problem instance, the same metric, and the same wall-clock budget. The Macaluso survey itself concedes the point: the authors conclude that no clear quantum advantage has been demonstrated for large-scale network optimization problems.

Hoefler, Häner and Troyer quantified the constraint in their 2023 paper in Communications of the ACM: quantum logical clock rates run at effective kilohertz to megahertz, against classical gigahertz. A quadratic speedup (which is what Grover-type search and most QAOA formulations offer) only wins at problem sizes so large that no practical telecom instance reaches them. No deployed telecom network is large enough to reach the crossover.

Most telecom problems are also data-heavy: loading a network traffic matrix or a subscriber database into a quantum processor costs at least N operations, erasing any sublinear or quadratic gain before the computation begins. This is the I/O problem, and it eliminates most of the advertised use cases before hardware limitations apply. For churn prediction, fraud detection, network-traffic anomaly detection, and large-scale capacity planning, loading the classical dataset into the processor requires at minimum N gate operations, an overhead that consumes whatever speedup the algorithm provides. Babbush and colleagues at Google reached the same conclusion in PRX Quantum in 2021: quadratic speedups are insufficient once error-correction overhead is included.

The problems that could in principle qualify are those with small-input, large-compute structure: a tiny problem encoding that produces an exponentially large intermediate computation. Quantum chemistry and materials simulation fit this pattern. No telecom optimization problem does. The closest candidate is single-cell MIMO maximum-likelihood detection, where the input is a small antenna-by-user matrix but the search space grows exponentially in users times modulation order. Kim, Singh, Venturelli, Jamieson and colleagues (X-ResQ, published at ACM MobiCom 2025) achieved near-optimal throughput for a 4×6 MIMO system with 16-QAM using 240 qubits on a D-Wave annealer via reverse annealing, with 220 µs annealer compute time.

Two practical blockers remain. The detection must run in sub-millisecond real time at a base station, and cloud-annealer latency cannot meet that requirement. The result has also not been compared against a well-tuned classical MMSE or sphere decoder at production scale.

Variational quantum algorithms face a separate structural problem. Barren plateaus (exponentially vanishing gradients as system size grows) undermine trainability at the scales where telecom problems become interesting. Larocca, Cerezo and colleagues reviewed the evidence in Nature Reviews Physics in 2025 and concluded that barren plateaus appear generically in deep variational circuits. Training variational quantum algorithms is NP-hard (Bittel and Kliesch, Physical Review Letters, 2021). Some structured ansätze evade barren plateaus, but those tend to be classically simulable, leaving no route to practical quantum advantage on telecom workloads.

Fault-tolerant gate-model advantage on telecom optimization is not this decade. The one gate-model certainty that telcos must act on is the threat: a cryptographically relevant quantum computer running Shor’s algorithm against RSA and ECC. Gidney’s 2025 estimate (arXiv:2505.15917), under surface-code assumptions with 10⁻³ gate error, puts that at fewer than one million noisy qubits in under a week. Telcos must therefore pursue PQC migration as a defensive programme.

Quantum annealing and the Ising machine question

If quantum computing for telecom optimization has a poster child, it is NTT DOCOMO’s work with D-Wave. In August 2024, DOCOMO announced a pilot (subsequently moved into production) of a quantum application in its mobile network: optimizing base-station tracking areas to reduce paging signals. D-Wave’s hybrid solver completed the task in 40 seconds against 27 hours for an unnamed general-purpose solver, with a 15% reduction in paging signals. A second application, announced jointly with D-Wave in August 2026, optimised tracking-area lists across a representative instance of 333 base stations, finishing in about five minutes; after operational apply, daily-peak location-registration signals fell 65.3% and paging signals fell 7.0%.

DOCOMO runs both in production with measurable operational savings, but neither is a quantum advantage demonstration. The comparison in both cases is against an unnamed solver, not a well-tuned classical heuristic matched on instance, metric and hardware generation. D-Wave’s hybrid solver is largely classical (the quantum annealer handles a subproblem) and the speedup claim conflates the entire hybrid pipeline with the quantum component. Every one of these results should be tested against a fair classical baseline before anyone claims quantum advantage.

This pattern recurs across every telco annealing result I have examined. Telecom Italia ran a proof-of-concept optimising Physical Cell Identifier (PCI) planning on a D-Wave 2000Q. The MIMO detection work used reverse annealing. Multiple teams have formulated RWA as QUBOs and run them on D-Wave hardware. In every case, the result is a valid proof-of-concept at small scale, and in every case the classical baseline is either unnamed, non-optimised, or absent.

The machines that telcos more often actually deploy are classical. The Fujitsu Digital Annealer is application-specific CMOS hardware (a dedicated DAU chip, fully-connected up to 100,000 variables in the third generation) that solves QUBOs. Telefónica, in a February 2025 partnership with the Government of Biscay, installed a Digital Annealer at a central office in the Basque Country – the first housed outside Japan, for multi-industry optimisation. Toshiba’s Simulated Bifurcation Machine runs on FPGAs and GPUs; Toshiba’s own materials report that the discrete SB algorithm solved a one-million-bit Ising problem in 30 minutes on 16 GPUs. NEC, Hitachi and NTT have their own classical Ising machines. These are powerful classical combinatorial solvers.

Benchmark comparisons between these platforms and D-Wave’s quantum annealer show no consistent quantum advantage. Oshiyama and Ohzeki compared D-Wave’s Hybrid Solver Service, Toshiba SBM, Fujitsu DA and PC-based simulated annealing across standard problem sets in Scientific Reports in 2022: no universal winner. D-Wave’s hybrid solver was best on MQLib instances, the Digital Annealer on NAE-3SAT, SBM on the Sherrington-Kirkpatrick model.

Should these classical machines be part of a quantum conversation? Only as the baseline. A telco evaluating combinatorial solvers for RWA or scheduling should benchmark Gurobi, CPLEX, the Digital Annealer, SBM and a well-tuned simulated annealing implementation on the same instance. If D-Wave’s hybrid solver wins on that benchmark, it earns its place. If the Digital Annealer wins, the telco has a better classical solver and does not need the word “quantum” attached to it.

QUBO formulation on classical annealers does not build as much transferable capability for future quantum hardware as vendors claim. It transfers to annealing and to QAOA – the modalities with barren plateaus and no demonstrated advantage. If fault-tolerant gate-model machines eventually produce a telecom result, there is no strong reason to think QUBO encoding is the path in. Only the data hygiene and problem-scoping work transfers to future hardware. The QUBO encoding is specific to annealing and QAOA and may not apply to gate-model architectures.

Quantum sensing and timing for telcos

Vendor decks rarely feature timing and synchronization, yet this front has procurable, production-grade equipment that telcos can deploy today.

5G TDD and positioning services require tight phase alignment, typically sourced from GNSS. GNSS is vulnerable to jamming and spoofing, threats that have moved from theoretical to operational, particularly in regions near conflict zones and in urban environments where inexpensive jammers are available online. When GNSS goes down, a base station’s local oscillator drifts, and synchronization degrades. For TDD networks, that drift translates directly into interference between uplink and downlink.

The solution is already standardised. ITU-T G.8272.1 defines the enhanced primary reference time clock (ePRTC), which disciplines a GNSS receiver with an always-on caesium atomic clock. When locked to GNSS, the clock maintains ±30 ns accuracy to UTC. During a GNSS outage, it holds within ±100 ns. The holdover window ranges from about 14 days in widely deployed baseline configurations up to 40 days depending on how long the clock was locked before the outage (the 2024 revision of G.8272.1 made this parametric). Oscilloquartz’s OSA 3350 in its SePRC+ variant is specified for ±100 ns holdover for 100 days guaranteed, 150 days typical. Microchip ships competing products. These are procurable, deployable, and operational today: atomic clocks using the hyperfine transition of caesium-133 atoms as a frequency standard. The engineering is mature, the ITU standards exist, and the products ship. For a telco spending time on “quantum strategy,” deploying ePRTC at core sites and evaluating chip-scale atomic clocks at edge sites would deliver more operational value this year than any optimization pilot.

Beyond timing, quantum sensing has two applications relevant to telecom infrastructure, both further out.

Rydberg atom RF receivers use electromagnetically induced transparency in alkali vapour cells to measure electric fields with SI-traceable, self-calibrating precision. They could serve as spectrum-monitoring sensors and, speculatively, as compact antennas. The physics is real: a June 2026 demonstration (arXiv:2606.25555) achieved 44.6 MHz instantaneous bandwidth, a substantial advance over earlier narrowband results. The limits are also real. Highest sensitivity occurs only near discrete atomic resonances. The instantaneous bandwidth is narrow compared with classical wideband receivers, and SWaP-C is not competitive with conventional antennas for most deployments. This is genuine research progress, roughly 5 to 10 years from field deployment.

Quantum gravity gradiometers could help telcos survey underground cable routes and utility corridors. Stray, Lamb, Kaushik and colleagues at the University of Birmingham published the first field demonstration in Nature in February 2022: a cold-atom dual-cloud gradiometer detected a 2-metre-wide utility tunnel buried about 1 metre below the surface, locating it to ±0.19 metres horizontally. The gradiometer configuration cancels common-mode vibration, which is the practical advantage over classical gravimeters that struggle outdoors. Directly relevant to fibre-route surveying and avoiding utility strikes when trenching. Roughly 3 to 7 years from routine commercial field use.

One clarification that belongs in any quantum-sensing discussion with telcos: distributed acoustic sensing (DAS) is classical, not quantum. It uses coherent optical time-domain reflectometry on standard fibre to detect vibration and strain. It is useful for intrusion detection, cable monitoring and seismic sensing but involves no superposition or entanglement. Operators frequently mistake DAS for quantum sensing; the error should be corrected whenever it appears.

QKD and quantum communications

Quantum key distribution (QKD) is the most commercially advanced quantum technology in telecom, and it is also the one where the gap between vendor positioning and security-agency guidance is widest.

The commercial deployments are real and growing. China Telecom reports more than 6.8 million quantum communication subscribers across more than 40 cities and describes quantum communication as the primary revenue driver within its quantum segment, with 65.4% revenue growth in 2025. China Telecom Quantum Group holds a controlling stake in QuantumCTek, which China Telecom describes as one of the few companies worldwide with large-scale industrialisation capability across quantum communication, computing and measurement. The Beijing-Shanghai backbone spans 2,032 km with 32 trusted relay nodes. The Pan Jianwei group’s integrated space-to-ground network adds 4,600 km across 700+ fibre segments and two satellite-ground links serving over 150 users.

Singapore’s NQSN+ programme, backed by IMDA, supports Singtel and SPTel in deploying nationwide quantum-safe networks. Singtel markets a hybrid PQC-plus-QKD Quantum-Safe Network as an enterprise service and won the Frost & Sullivan 2025 Singapore Quantum Safe Network Service Provider award. SPTel partnered with SpeQtral, Toshiba and ST Engineering to trial QKD on its diverse fibre network.

In April 2022, BT and Toshiba connected a three-node quantum-secured fibre ring across London over Openreach private fibre, with EY as the initial commercial customer. Toshiba published a twin-field QKD field trial in Nature in April 2025: coherent key distribution over 254 km of deployed commercial fibre in Germany (Frankfurt–Kehl, access via GÉANT) using semiconductor avalanche photodiodes at room temperature. The trial eliminated the cryogenic requirement that had confined earlier coherent-QKD work to the laboratory.

ID Quantique has delivered national-scale networks in Poland (1,770 km connecting five HPC centres) and Slovakia (combining QKD with post-quantum cryptography for government communications). EuroQCI is building an EU-wide quantum communication infrastructure with terrestrial and space segments; ESA and the European Commission signed an implementation agreement in January 2025; the space-segment prototype (Eagle-1) is targeted for late 2027, with full operational EuroQCI a later integration milestone.

India’s C-DOT launched fourteen indigenous quantum products in August 2026, covering QKD hardware, PQC-based encryptors and specialised communication components, explicitly framed as a dual-track architecture: physics-based QKD for critical communications alongside algorithmic PQC across enterprise and defence networks.

SK Telecom has deployed QKD devices on its backbone and offers quantum-safe VPN services. SK Telecom and Samsung partnered with ID Quantique to ship the Galaxy A Quantum in May 2020, the world’s first QRNG smartphone with a 2.5 mm × 2.5 mm Quantis chip integrated into the Android Keystore.

QKD revenue is growing, subscribers number in millions, and national backbones span thousands of kilometres. The security objections are equally documented.

Six Western security bodies have published formal positions discouraging QKD as a standalone solution. NSA recommends post-quantum cryptography as a more cost-effective and easily maintained solution and cites five specific QKD limitations: no source authentication, special-purpose equipment requirements, increased infrastructure cost and insider-threat risk, difficulty of validation, and denial-of-service risk. CNSA 2.0 formally excludes QKD for DoD and National Security Systems. NCSC will not support QKD for government or military use and endorses PQC as the primary defence.

ANSSI, BSI, the Dutch NCSA and the Swedish NCSA jointly published a position paper in January 2024 concluding that QKD is usable only in niche cases, is not yet sufficiently mature from a security perspective, and that PQC migration should be the clear priority.

The map of QKD commercialisation is close to the inverse of the map of security-agency scepticism. QKD sells where the state is the anchor customer and sovereignty is part of the specification: China, Singapore, South Korea, India, EU member states building EuroQCI. Where the buyer follows NSA or NCSC procurement guidance, QKD does not sell; the commercial adoption aligns with state industrial-policy investment, a dynamic that “unhackable communications” and “rejected technology” both miss.

For a telco, the viable commercial models are narrow. A government-anchored managed service where a national programme pays for the infrastructure. An enterprise differentiator bundled with PQC for regulated banks. Sovereignty positioning against foreign vendor supply chains. Singtel’s hybrid PQC-plus-QKD model is the commercially honest version.

Long-range fibre QKD today requires trusted relay nodes, where keys are exposed at each node. Huttner and colleagues established in npj Quantum Information in 2022 that long-range QKD without trusted nodes is not possible with current technology. Quantum repeaters (which would remove the trusted-node requirement) need quantum memories that can store entangled states, and quantum memory technology remains in the laboratory. The EuroQCI planning documents cite repeater demonstrations as a future milestone. Until repeaters mature, any continental-scale QKD network depends on relay nodes where keys exist in the clear, and a single compromised relay exposes those keys without breaking the physics.

QRNG (quantum random number generation) is the most mature and least controversial quantum communications product. It is procurable, independently validated, shipping in handsets, SIM cards, HSMs and 5G-core authentication systems. It is a genuine, low-cost enhancement to key generation, and it works without the infrastructure requirements or the security-agency debate that attend QKD.

PQC migration

Every telco that handles encrypted traffic, which is every telco, faces a cryptographic migration driven not by when a CRQC will arrive but by regulatory deadlines that are already fixed. In the United States, EO 14412 directs federal agencies to transition high-value assets and high-impact systems to PQC key establishment by 31 December 2030 and digital signatures by 31 December 2031, with separate FAR rulemaking directed for the contractor track. CNSA 2.0 requires ML-KEM (formerly CRYSTALS-Kyber) and ML-DSA (formerly CRYSTALS-Dilithium) for national security systems on an accelerating timeline.

The EU Commission has recommended that member states create quantum-safe migration roadmaps. Singapore’s Monetary Authority (MAS) has announced supervisory expectations, due later in 2026, targeting quantum resilience for financial institutions by the end of the decade.

The GSMA Post-Quantum Telco Network Taskforce, launched in September 2022 by IBM and Vodafone and now with more than 50 member companies including 20+ major operators, has published the Post-Quantum Telco Network Impact Assessment, PQC Guidelines for Telecom Use Cases, and PQ.07 on PQC for non-terrestrial networks. The work is real and the publications specific. The urgency comes from regulatory obligations. Debating Q-Day probabilities is close to irrelevant when the deadlines are set.

For telcos, the migration touches every layer: signalling protocols, SIM authentication, VPNs, TLS termination, software update signing, inter-operator connections, subscriber data encryption. Building a cryptographic inventory (a systematic record of where cryptography secures the network, the devices and the systems) is the prerequisite. Crypto-agility (the ability to swap cryptographic algorithms with minimal disruption) is the architecture goal. NIST published the final PQC standards (FIPS 203 ML-KEM, FIPS 204 ML-DSA, FIPS 205 SLH-DSA) in August 2024, and FN-DSA is expected as a subsequent standard. ML-KEM and ML-DSA run on conventional processors. The migration is therefore a software, procurement and governance programme: large, complex and expensive, but executable with existing infrastructure.

Adversaries can intercept and store encrypted data today, then decrypt it once a CRQC exists. Harvest now, decrypt later is the threat model that makes the migration urgent independently of the CRQC timeline: any secret with a shelf life beyond the plausible CRQC window is already at risk.

What a telco should do now

Immediately. Deploy ePRTC (ITU-T G.8272.1) with caesium holdover at core sites and evaluate chip-scale atomic clocks (CSAC) and fibre time transfer at the edge. CSAC is a different product class: holdover of days rather than months, at a fraction of the cost and size, suited to bridging GNSS outages at cell sites rather than replacing core-site caesium clocks. This addresses a real, present GNSS jamming and spoofing risk with procurable equipment.

Start PQC migration: build the cryptographic inventory, assess crypto-agility, and begin implementing ML-KEM and ML-DSA where the NIST standards and vendor support allow. Treat QRNG as a low-cost enhancement for key generation in HSM, SIM and 5G core deployments.

Near-term (1–3 years). Benchmark quantum-inspired classical Ising machines (Fujitsu Digital Annealer, Toshiba SBM) against Gurobi, CPLEX and well-tuned simulated annealing on your own network-planning problem instances. Adopt them where they win on the same benchmark, the same metric and the same wall-clock budget – and call them what they are: better classical solvers. Run quantum annealing (D-Wave hybrid) as R&D only, with published apples-to-apples classical baselines. Pilot QKD only for specific high-value point-to-point links where trusted-node exposure is acceptable and there is regulatory or customer demand; always pair it with PQC and classical authentication.

Watch-and-wait (3–10 years), with trigger milestones. Track quantum gravity gradiometers for fibre-route utility surveying (trigger: a commercial field instrument at civil-engineering price points). Track Rydberg RF receivers for spectrum monitoring (trigger: bandwidth and SWaP-C parity with classical antennas). Track quantum repeaters (trigger: a repeater link across a metro network that removes trusted nodes). Track fault-tolerant logical-qubit counts and, critically, any published fault-tolerant resource estimate showing a super-quadratic speedup on a real telecom problem. Absent that estimate, assume no gate-model quantum-computing advantage for telecom this decade.