The Quantum Internet

In short

A quantum internet is a global network in which distant nodes share not just classical bits but entanglement — Bell pairs, larger entangled states, and the ability to teleport arbitrary qubits between them. The defining 2018 paper by Stephanie Wehner, David Elkouss, and Ronald Hanson (Science 362, 6412) lays out a six-stage roadmap from today's trusted-node QKD networks up to fully networked fault-tolerant quantum computers. Each stage unlocks qualitatively new capabilities: stage 1 gives you ISRO-style key distribution with trusted relays; stage 3 (entanglement distribution) enables device-independent QKD and some leader-election protocols; stage 4 (quantum memory) enables blind quantum computing; stage 5 (error-corrected qubits) enables distributed Shor's algorithm and high-precision long-baseline interferometry; stage 6 (networked fault-tolerant computers) is the endgame. The core protocols are the ones you have already met: entanglement swapping (ch.52) as the repeater primitive, quantum teleportation (ch.50) as the end-to-end qubit-transfer mechanism, distillation (ch.entanglement-as-resource) to clean up noisy pairs, and BB84 (ch.152) as the killer application. As of 2026, the world is at stage 1 (trusted-node deployments in China, Europe, India) with stage 2–3 prototypes in laboratory and satellite form. India's National Quantum Mission targets stage 3 for intercity links by 2032.

The classical internet moves bits. You type a message, it is encoded as a sequence of zeros and ones, these are sent as electrical or optical pulses down fibres, they are amplified at repeaters, they are received at a destination, and the message reconstructed. This architecture has conquered the planet. Google Search, UPI, WhatsApp, IRCTC — all of them ride on a single abstraction: transporting classical information from A to B.

A quantum internet is different in exactly one way. It transports entanglement. Two distant nodes, at the end of a quantum-internet operation, share a Bell pair — a pair of qubits in the correlated state |\Phi^+\rangle = \tfrac{1}{\sqrt 2}(|00\rangle + |11\rangle) — and can, using that Bell pair plus normal classical communication, do things no classical network can do.

This is not a faster internet. It is a separate infrastructure layered alongside the classical one. The classical internet is not going away; it remains the carrier for all ordinary traffic. The quantum internet is the specialised channel for cryptographic key exchange, for distributed quantum-computing workloads, for long-baseline telescope synchronisation, for secure multi-party computation that exploits quantum advantage. It is an analogue of the shift from postal mail to email — the new channel doesn't replace the old; it creates new possibilities entirely.

The question this chapter answers is: what, concretely, does a quantum internet let you do? Where are we on the way to building one? What is the engineering gap between today's ISRO Bengaluru–Mt. Abu demonstration and a full continental quantum network? The answers come from the six-stage roadmap published by Wehner, Elkouss, and Hanson in 2018 — the clearest framing of quantum-internet capabilities that the field has produced, and the one that every national quantum programme, including India's, uses as its reference.

Classical versus quantum internet — what changes

Here is the structural difference in one picture.

Classical internet: transport bits. Quantum internet: transport entanglement (plus classical bits alongside, for protocols that need them). The quantum internet does not replace the classical one; it layers on top.

What you do with the entanglement is where the action lives. Three primitives you have already met are the moves:

Quantum teleportation (ch.50): given a shared Bell pair plus two classical bits, transfer an arbitrary qubit state |\psi\rangle from Alice to Bob.
Superdense coding (ch.51): given a shared Bell pair, Alice can send Bob two classical bits by transmitting only one qubit.
BB84 key distribution (ch.152): given a photon-by-photon quantum channel (not even Bell pairs — just non-orthogonal single qubits), Alice and Bob can build a provably secure shared key.

And with entanglement in hand, you get higher-order protocols: blind quantum computing (Alice delegates a quantum computation to Bob's quantum computer without Bob learning what the computation was), anonymous leader election over a multi-node quantum network, device-independent QKD (security does not even require trusting the hardware, just the classical statistics of measurement outcomes), quantum Byzantine agreement, clock synchronisation below the classical Heisenberg limit, and eventually networked fault-tolerant quantum computing — in which a single logical qubit is spread across multiple quantum computers, each too small to do the full job alone.

None of these applications is a nice-to-have incremental improvement over its classical counterpart. Each of them is impossible on a classical network, full stop, by the same arguments that make quantum teleportation impossible classically: a classical network cannot transport quantum correlations, because classical correlations are bounded by local realism (Bell's theorem) and quantum correlations are not.

The Wehner-Elkouss-Hanson six-stage roadmap

Building the infrastructure is the hard part. Bell pairs at scale need entanglement generation, routing, memories, distillation, error correction, and quantum-network stacks analogous to TCP/IP. Different levels of infrastructure unlock different capabilities. The 2018 Wehner-Elkouss-Hanson paper (Quantum internet: A vision for the road ahead, Science 362, 303) lays out six named stages, each defined by the network's technical capability, and each unlocking specific applications.

The Wehner-Elkouss-Hanson six-stage roadmap. Each stage unlocks specific applications and requires specific infrastructure. As of 2026, stage 1 is widely deployed; stages 2–3 are in laboratory and satellite demonstration; stages 4–6 are research horizons.

Here is each stage in the detail the roadmap prescribes.

Stage 1 — Trusted repeater QKD. The lowest tier. Each pair of adjacent nodes runs standard BB84; an end-to-end key is assembled by classically relaying keys through intermediate nodes, which must be trusted (a compromised intermediate node compromises the whole chain). This is what China's Beijing–Shanghai QKD backbone does, and what India's metro-scale QKD pilots are deploying. The ISRO 2022 Bengaluru–Mt. Abu satellite-QKD demonstration is a variant: the satellite is the "trusted relay" between the two ground stations. This stage requires no Bell pairs and no quantum memories; it is the minimal viable quantum internet.

Stage 2 — Prepare-and-measure. End-to-end QKD without trusted relays. Alice prepares photons; Bob measures them; the photons travel through a fibre that may include non-trusted classical amplifiers (or pure passive routing). Achieves BB84 with security against any eavesdropper, not just against untrusted relays. Requires per-photon survival over the whole distance — which is the thing quantum repeaters are needed to provide. Stage 2 without repeaters is limited to about 500 km in fibre.

Stage 3 — Entanglement distribution. Distant nodes can share Bell pairs at demand, on network timescales. Requires full quantum-repeater infrastructure (entanglement generation, memory, swapping, distillation). Unlocks device-independent QKD (DIQKD), in which the security proof assumes only that the classical measurement statistics violate a Bell inequality — meaning you do not need to trust the hardware itself. This is a qualitative security upgrade. Stage 3 also supports leader election (choosing a random leader from n parties with provable fairness), fundamentally different from classical randomness beacons. India's NQM targets stage 3 for intercity links by 2032.

Stage 4 — Quantum memory network. Each node has a quantum memory that can store a qubit for longer than a round-trip classical communication. Enables blind quantum computing (Alice delegates a computation to Bob's quantum server without Bob learning the computation); anonymous voting with cryptographic unlinkability; secure multi-party computation in which no party learns anything beyond the function value. Requires memories with T_2 coherence times of seconds, not microseconds — an active research frontier.

Stage 5 — Few-qubit fault tolerance. Each node can store and manipulate a small number of error-corrected (logical) qubits. Enables distributed quantum algorithms in which a single logical qubit is spread across multiple nodes — the beginning of networked Shor's algorithm, for example. Enables quantum-enhanced telescopy: long-baseline interferometry with entanglement-distributed reference clocks, achieving angular resolution that no single-site telescope can match. Enables clock synchronisation at precision beyond the classical Heisenberg limit, essential for next-generation GPS.

Stage 6 — Quantum computing networks. Full networks of fault-tolerant quantum computers, running distributed quantum algorithms with logical-qubit passing between machines. This is the endgame: a global quantum-computing resource, analogous to today's cloud compute but for quantum workloads. Timeline: the community's rough consensus is 2040 at the earliest, possibly later. The bottleneck is simultaneously scaling up every building block.

Progress through the stages is not a straight line. A stage-4 system in one laboratory can coexist with stage-1 systems in deployment; the roadmap is about the most advanced service the network can provide, not a uniform upgrade. The picture in 2026 is a patchy mosaic: stage 1 in operational deployment, stage 2–3 in satellite demonstrations and university labs, stages 4+ confined to individual small-scale research systems.

Key protocols — the plumbing

Three protocols from earlier chapters are the stitching that holds a quantum internet together. A quick tour:

Entanglement swapping (ch.52) is the repeater primitive. Two short-hop Bell pairs meeting at an intermediate node are converted, via a Bell measurement on the two inner qubits, into one long Bell pair between the outer nodes. The procedure does not require any qubit to traverse the long distance — only classical bits do. This is the move that lets a quantum internet span continents without any single photon surviving the full trip.

Entanglement distillation (ch.entanglement-as-resource) takes many noisy Bell pairs and produces fewer cleaner pairs. Necessary because every physical operation is noisy, and without distillation, fidelity decays rapidly as you chain swaps. Protocols like DEJMPS (Deutsch, Ekert, Jozsa, Macchiavello, Popescu, Sanpera, 1996) use CNOT and measurement on two input pairs to produce one output pair of higher fidelity with probability.

Quantum teleportation (ch.50) is the end-to-end qubit-transfer mechanism. Given a shared Bell pair plus two classical bits, Alice can send Bob any qubit state — including the halves of larger entangled states, which is how distributed quantum computing transmits its intermediate results. Teleportation is both a primitive in its own right and a tool for larger protocols.

Routing, addressing, and the quantum network stack are where the field is still figuring out the right abstractions. The classical internet stack (link, network, transport, application — the TCP/IP stack) has rough quantum analogues but needs specific adaptations: how do you route a Bell-pair request across a network with finite memory lifetimes? How do you reconcile multiple requests competing for the same path? How do you signal errors? A full quantum-network protocol stack (the "QIS2" proposal from Delft, analogous to TCP/IP) is an active topic, with prototypes being tested in Europe's Quantum Internet Alliance.

The physical layer — fibre, free space, satellite

Every quantum-internet deployment is built from two (or three) kinds of links.

Fibre links use standard telecom single-mode fibre. Loss is 0.2 dB/km at 1550 nm. Feasible for metro-scale (up to roughly 100 km per segment) and, with repeaters, for intercity. The advantage: existing fibre infrastructure can often be reused. The disadvantages: exponential loss, and the engineering need for full repeater chains for continental distances.

Free-space links use line-of-sight laser communication through the atmosphere. Loss is a few dB plus geometric beam spread. Feasible between rooftops in a city or up to a satellite. The advantage: compatible with satellites and with locations where fibre is unavailable. The disadvantages: weather dependence, strict pointing requirements, and the sub-kilometre range in pure ground-to-ground deployments.

Satellite links use a spacecraft as a relay or an entangled-photon source. China's Micius (2016–) is the flagship; India's ISRO 2022 Bengaluru–Mt. Abu demonstration, the CV-QKD Ahmedabad trials, and the upcoming QSAT-01 mission (planned launch window 2027–2028) are the Indian contributions. Satellites dominate for intercontinental distances; ground repeaters will dominate for within-country intercity links; both coexist in the eventual deployed quantum internet.

Applications beyond QKD

It bears repeating: the quantum internet is not mainly about QKD. QKD is the only application that is ready for deployment today, which is why roadmaps start there — but the long-term value is in the stage-4-and-above applications, where capabilities appear that have no classical analogue.

Blind quantum computing (stage 4). Alice has a computation she wants to run on a powerful quantum computer owned by Bob, but she does not want Bob to learn what the computation is. Classical blind computing (homomorphic encryption) is possible but has enormous overhead. Quantum blind computing, via a protocol of Broadbent, Fitzsimons, and Kashefi (2009), has near-linear overhead and cryptographically provable blindness. It requires a quantum-memory-capable link between Alice and Bob.

Long-baseline telescopy (stage 5). The Event Horizon Telescope used classical interferometry across nine sites to image Sgr A*. Quantum-assisted interferometry — distributing entanglement between telescope stations so that single photons from an astronomical source can be correlated coherently — promises angular resolution that goes as 1/\lambda D with D the baseline, regardless of baseline length. With a global network of telescopes, this is radio-astronomy-quality resolution at visible wavelengths.

Clock synchronisation (stage 5). Networks of atomic clocks distribute a common time reference by exchanging timing signals. Classical synchronisation is limited by photon-counting shot noise. With shared entanglement, the ultimate Heisenberg-scaling of 1/N (rather than the classical 1/\sqrt N) is achievable. Essential for next-generation GPS, for LIGO-scale gravitational-wave detectors, and for any distributed-clock application.

Distributed quantum computing (stage 5+6). A 10-million-qubit fault-tolerant quantum computer is the target for cryptographically-relevant Shor's algorithm. Building one monolithic 10M-qubit machine is hard. Building a network of smaller machines — each 100k-qubit, connected by entanglement links — is a different engineering problem, possibly easier. The "modular quantum computer" research direction (IBM's modular roadmap, Amazon Braket's QPU interconnect studies) is betting on this.

Quantum Byzantine agreement and anonymous multi-party protocols (stage 4+). Classical consensus protocols (like blockchain PoW) need heavy computation for Byzantine tolerance. Quantum-assisted Byzantine agreement with pre-shared entanglement achieves fault-tolerant consensus with round-complexity and communication-complexity advantages proven in 2002 by Fitzi, Gisin, and Maurer.

Example 1 — stage assessment for the 2022 ISRO Bengaluru–Mt. Abu QKD experiment

Setup. In 2022, ISRO's Space Applications Centre (Ahmedabad) demonstrated free-space quantum key distribution between Bengaluru and Mt. Abu using an in-orbit relay. The demonstration delivered a shared BB84-like secret key between the two ground stations. You are asked to place this experiment on the Wehner-Elkouss-Hanson roadmap.

Step 1 — identify the protocol used. The experiment was a form of BB84 with the satellite acting as a trusted relay: the satellite measured photons from one ground station, generated a new key, and separately distributed a correlated key to the other ground station. The satellite itself had to be trusted — if someone compromised the satellite, they could read the key.

Why this is trusted-relay: the satellite does not distribute entangled photon pairs across the 1100 km ground-to-ground distance. It distributes single photons in short bursts to each ground station separately, and the classical-key reconciliation is done by the satellite. So both ground stations are secure only to the extent that the satellite is secure.

Step 2 — match the protocol to a stage. Wehner-Elkouss-Hanson stage 1 is defined as: "Trusted repeater — only the repeaters (classical nodes) have quantum functionality; any trusted node can read the key." The ISRO 2022 setup fits this definition exactly: the satellite is the trusted repeater, and each ground-satellite link is a short prepare-and-measure BB84 exchange.

Step 3 — what would need to change to reach stage 2? Stage 2 requires end-to-end prepare-and-measure: Alice sends photons directly to Bob (via passive relays, not trusted relays), and the security is end-to-end. A satellite performing entanglement distribution — sending one photon of a Bell pair to each ground station, rather than independent single photons to each — would move the setup to stage 3. Micius (2017) did a distribution experiment at 1200 km, placing it at stage 3.

Step 4 — what would need to change to reach stage 3? Stage 3 requires sustained entanglement distribution at network timescales, not single demonstrations. For ISRO, this means routine entanglement distribution on every satellite pass, with integration into a multi-node Indian network spanning at least Bengaluru, Mt. Abu, Delhi, and Mumbai. The NQM's 2032 target aligns with this.

Result. ISRO 2022 = stage 1 (trusted-relay QKD with satellite as relay). The experiment is a landmark for India as a proof-of-concept that satellite-QKD is technically feasible on Indian infrastructure; the trajectory to stage 3 is the Mission's 10-year horizon.

What this shows. Placing a concrete deployment on the six-stage roadmap requires separating what the hardware did from what a more advanced version could do with the same hardware. The Micius-style entanglement distribution is stage 3, but the first ISRO demonstration was closer to Micius's earlier trusted-node work (2017–2018) rather than its later entanglement-distribution milestones. Progress through stages is cumulative: you can demonstrate stage 3 with a stage-1 deployment simply by changing the mode of the satellite's photon source.

Example 2 — distributed VQE between two QPUs via the quantum internet

Setup. Two quantum processing units (QPUs) sit in two different labs, each with n qubits and connected by a quantum-internet link capable of on-demand Bell-pair distribution. You want to run a variational quantum eigensolver (VQE) on a 2n-qubit problem — say, computing the ground-state energy of a molecule that is too large for either QPU alone. You need to execute a 2n-qubit unitary U(\theta) with a variational parameter \theta, optimise \theta to minimise expected energy, and return the final energy. The question: how much entanglement do you consume per VQE iteration?

Step 1 — identify the non-local gates. In any VQE ansatz, most gates are local to a single qubit or a single QPU. The expensive part is the two-qubit gates that cross the partition between the two QPUs. Call this count G_{\mathrm{cross}}. For a reasonable hardware-efficient ansatz with depth d on 2n qubits, G_{\mathrm{cross}} \approx d \cdot (n/2) (a constant fraction of the two-qubit gates cross the partition at each depth layer).

Step 2 — how to execute a cross-QPU two-qubit gate. The standard trick is gate teleportation: a CNOT between a qubit on QPU A and a qubit on QPU B can be executed using one Bell pair, two local CNOTs, one local Hadamard, two measurements, and two classical bits of communication. (Eisert, Jacobs, Papadopoulos, Plenio 2000.) So each cross-QPU CNOT costs 1 Bell pair + 2 classical bits.

Why one Bell pair per cross-gate: the gate-teleportation protocol uses a shared Bell pair as an "entanglement resource" that is consumed by the protocol. You cannot reuse the Bell pair for a second cross-gate — it is destroyed by the measurement step. Each additional cross-gate consumes another Bell pair.

Step 3 — Bell-pair budget per VQE iteration. One VQE iteration = one run of the ansatz U(\theta) + one energy measurement. The ansatz consumes G_{\mathrm{cross}} \approx d \cdot (n/2) Bell pairs. For n = 50 qubits per QPU and d = 100 depth layers, that is

G_{\mathrm{cross}} \approx 100 \cdot 25 = 2500 \text{ Bell pairs per iteration}.

A VQE optimisation typically needs 10^3 to 10^5 iterations. Call it 10^4 iterations. Total Bell pairs per VQE run:

G_{\mathrm{total}} = 10^4 \cdot 2500 = 2.5 \times 10^7 \text{ Bell pairs}.

Step 4 — rate requirement and stage. If you want to finish the VQE run in one hour (reasonable for a chemistry calculation), the required Bell-pair rate is

R = \frac{2.5 \times 10^7}{3600 \text{ s}} \approx 7000 \text{ Bell pairs per second}.

Continuous high-fidelity Bell-pair distribution at 7000 Hz between two distant QPUs requires a stage-5 quantum internet — error-corrected Bell pairs of fidelity above 99%, distributed at near-memory-bandwidth rates. This is beyond what today's networks can do; it is the engineering target for the 2040 horizon.

Result. A single VQE iteration between 50+50 qubits needs thousands of Bell pairs, and a full optimisation needs tens of millions. Running distributed VQE at useful scale is firmly a stage-5 capability.

What this shows. The Bell-pair cost of distributed quantum computing is steep. Each cross-QPU two-qubit gate is one Bell pair, and real algorithms have many such gates. This is the reason distributed QC is a stage-5+ application: you need fast, clean, continuous entanglement distribution, not occasional heroic-demonstration Bell pairs. The cost also suggests architectural choices — partitioning the problem to minimise cross-QPU gates (like a compiler pass in classical distributed computing), choosing ansätze with low G_{\mathrm{cross}}, and co-locating QPUs to reduce the per-Bell-pair latency. The quantum internet makes distributed QC possible, but the engineering work is in making it efficient.

Common confusions

"A quantum internet replaces the classical internet." No. It layers on top. The quantum internet needs classical channels alongside every quantum channel — for teleportation corrections, for key reconciliation, for routing control. The classical internet provides that alongside channel; there is no quantum replacement for it, and there does not need to be one. The quantum internet is a specialised overlay, not a substitute.
"Quantum internet = faster data transfer." No. Bell-pair distribution rates are, and will remain, far slower than classical-bit rates. A fibre carries gigabits per second of classical data; a quantum-internet link delivers kilohertz-to-megahertz of Bell pairs in the near term, possibly reaching gigahertz in the distant future. The quantum internet is about capabilities, not speed. Use the classical internet for YouTube; use the quantum internet for cryptographic keys.
"Stage 3 means the problem is solved." Stage 3 is the beginning of utility, not the end. It is where the first applications beyond QKD (device-independent QKD, leader election) become possible. The major research-horizon applications — blind computing, distributed algorithms, long-baseline telescopy — all need stages 4 or higher.
"QKD is obsolete once we have post-quantum cryptography." No. Post-quantum classical cryptography (lattice-based, hash-based) provides computational security against future quantum attacks. QKD provides information-theoretic security — security that holds against any computational adversary, bounded only by the laws of physics. They are complementary: post-quantum for bulk encryption, QKD for the most sensitive long-lived keys. Both will be deployed.
"Entanglement can send information faster than light." No. The no-communication theorem proves that shared entanglement alone transfers no information. Every quantum-internet protocol requires a parallel classical channel running at sub-light speed. The quantum internet is not faster-than-light; it is different-than-classical, which is a strictly stronger and more useful property.
"The quantum internet is 50 years away." Deployment of stage 1 is already here (2026). Stage 2–3 prototypes are operational in laboratories. Stage 4 is the 2030s. Stages 5–6 are the 2040s. The timeline is aggressive but not fantastical — the scale of engineering resembles classical telecom's 20th-century buildout, and the national quantum missions (US, China, EU, India, Japan) are collectively funding at a level comparable to 1970s telecom R&D.

Going deeper

You now know the definition of a quantum internet, the core protocols (swapping, distillation, teleportation), the Wehner-Elkouss-Hanson six-stage roadmap, the applications at each stage, the physical-layer options (fibre, free space, satellite), and how the 2022 ISRO demonstration places on the roadmap (stage 1). The sections below cover the quantum network stack, the NetQASM control-plane proposal, the distributed-computing theorems that quantify quantum advantage in communication complexity, and India's specific strategic positioning.

The quantum network stack

A working quantum internet needs a protocol stack with well-defined layers, analogous to the TCP/IP stack. The Dahlberg-Skrzypczyk-Coopmans-Wubben et al. (2019) proposal defines five layers:

Physical layer — photons, fibres, detectors, lasers. Analogous to classical layer 1.
Link layer — turning a physical link into a usable entanglement-generation service between two adjacent nodes. Heralded Bell-pair generation, MAC-style contention resolution when multiple qubits are queued.
Network layer — routing entanglement across multiple hops. Entanglement swapping at intermediate nodes. Analogous to classical IP.
Transport layer — end-to-end protocols for reliable entanglement delivery. Distillation scheduling, error handling, flow control.
Application layer — QKD, teleportation, blind computing, distributed algorithms.

The analogy to TCP/IP is strong but not perfect: the quantum link layer has no analogue of "bit errors" (measurement collapses them to definite states); the network layer has to handle entanglement lifetime as a first-class quantity, unlike classical packets which age without consequence.

A practical control plane is under development. NetQASM (Dahlberg et al., 2022) is a proposed low-level assembly language for quantum-network operations, analogous to RDMA-style primitives in classical data centres. A QKD application compiles down to NetQASM instructions that the network runtime executes. The European Quantum Internet Alliance's testbed runs an early NetQASM implementation.

Distributed computing speedups — the theorems

Communication complexity asks: how many bits must two parties exchange to compute a function of their joint inputs? Classical communication complexity has well-known lower bounds (e.g., \Omega(n) bits to decide whether Alice and Bob's n-bit strings are equal, in the deterministic worst case). Quantum communication complexity sometimes needs exponentially fewer qubits to solve the same problem.

The Buhrman-Cleve-Wigderson (1998) theorem gives one example: a function where classical bit-exchange requires \Omega(\sqrt n) bits but quantum qubit-exchange requires only O(\log n) qubits. The Raz (1999) theorem gives a stronger separation: a partial function with classical \Omega(n^{1/3}) versus quantum O(\log n). These are theoretical results establishing that quantum communication can be exponentially more efficient than classical for specific problems.

Practical implications are limited — the problems with large quantum communication advantage tend to be artificial — but the theorems establish the principle that the quantum internet's bandwidth is not the right metric. A quantum internet can solve problems that no classical network can, regardless of bandwidth.

Another landmark: Fitzi-Gisin-Maurer (2002) on Byzantine agreement. Classical Byzantine agreement needs \Omega(n^2) messages and requires cryptographic assumptions (signatures) to tolerate more than n/3 traitors. With pre-shared entanglement, the protocol is deterministic, tolerates up to n/3 traitors without cryptographic signatures, and has lower message complexity.

India's National Quantum Mission — networking pillar

The National Quantum Mission (NQM), launched in 2023 with ₹6003 crore over 8 years, identifies quantum communication as one of four verticals (the others being computing, sensing, and materials). The networking targets are:

By 2028: metro-scale fibre-based QKD networks in Delhi, Mumbai, Bengaluru, Hyderabad — each a fully trusted-node network of 5–10 nodes across the metropolitan area. (Stage 1.)
By 2030: at least two metros connected by trusted-node intercity QKD over fibre. Intercity distances ~1000 km. (Stage 1 at scale.)
By 2032: satellite-relayed QKD between any two Indian metros, routine operation. Extending the 2022 Bengaluru–Mt. Abu demonstration. (Stage 1 via satellite; stage 3 for the actual entanglement distribution step that the satellite performs.)
By 2035–2040: first Indian stage-3/4 prototype — a three-node entanglement-distribution network with quantum memories.

Key participating institutions: ISRO Space Applications Centre (Ahmedabad), Raman Research Institute (Bengaluru), TIFR (Mumbai), IIT Madras, IISc Bangalore, IIT Bombay, IIT Delhi, and industry partners QNu Labs (Bengaluru, quantum-safe cryptography), HCL Tech, TCS Research. The International Quantum Communication Infrastructure partnership with the EU and Japan gives Indian researchers access to the European Quantum Internet Alliance testbed.

The strategic logic is explicit: Aadhaar authentication (over 1.3 billion identities), UPI payments (₹17+ lakh crore per month as of 2025), the stock exchanges (₹4+ lakh crore daily), and inter-ministry secure communications all rely on RSA/ECC-based public-key cryptography that Shor's algorithm will eventually break. A two-layer defence — post-quantum classical cryptography (for bulk use) plus QKD (for the most sensitive long-lived keys) — is the operational plan. The quantum internet is, from this angle, a cryptographic hedge against 2035–2045 quantum-computing milestones.

Why the roadmap matters

The roadmap is useful because it gives a common language across disciplines. A hardware engineer working on atomic-ensemble memories, a theorist proving distillation rate bounds, a cryptographer designing device-independent QKD protocols, and a government minister funding the whole thing can all agree on which stage they are targeting without arguing about definitions. That shared vocabulary is not trivial in a field where "quantum internet" has been a fuzzy buzzword for two decades.

The Wehner-Elkouss-Hanson paper is also honest about what the roadmap is not. It is not a schedule: stages are not dated, because dating any of them (beyond stage 1, which is already deployed) is speculation. It is not a universal ordering: a particular deployment might skip stage 2 and jump to stage 3, or achieve stage 4 capabilities for one application while still at stage 2 for another. The roadmap is a capability taxonomy, nothing more.

Used correctly, it answers the question the reader started this chapter with: what, concretely, does a quantum internet let you do? Stage-by-stage, the answer is crisp. And it places the present moment — ISRO 2022 in India, Micius 2017 in China, SECOQC 2008 in Vienna, the 2024 Delft three-node memory network, the various 2025–2026 metro QKD deployments — on a clear trajectory toward capabilities that are qualitatively new, not incremental.

Where this leads next

Quantum Repeaters — the hardware that makes stage 3 and above possible.
BB84 Protocol — the stage-1/2 killer application.
A Tiny Quantum Network — the minimal three-node case that scales to the full internet.
Entanglement Swapping — the core primitive every repeater node runs.
Quantum Teleportation — the end-to-end qubit-transfer mechanism.
Distributed Quantum Computing — the stage-5+ application where Bell-pair consumption becomes the dominant resource cost.

References

S. Wehner, D. Elkouss, R. Hanson, Quantum internet: A vision for the road ahead (Science 362, 303, 2018) — DOI link / arXiv preprint. The canonical six-stage roadmap paper.
H. J. Kimble, The quantum internet (Nature 453, 2008) — arXiv:0806.4195. The original vision paper that launched the field.
A. Dahlberg et al., A link layer protocol for quantum networks (SIGCOMM 2019) — arXiv:1903.09778. The Delft quantum-network-stack proposal.
S.-K. Liao et al., Satellite-relayed intercontinental quantum network (Phys. Rev. Lett. 120, 2018) — arXiv:1801.04418. The Micius intercontinental QKD result.
Wikipedia, Quantum network — concise overview with comprehensive links to current deployments.
Government of India, National Quantum Mission — the mission document laying out India's quantum-communication roadmap through 2031.